Decentralization country/continent code: reduce duplication and redundant node_ip_classification reads

[2ndtlmining]

Two related nits from the ultrareview of PR #145/#147 (issue #138's country/continent work), not fixed yet since they're efficiency/maintainability, not correctness:

1. Redundant full-table reads per snapshot cycle

snapshotManager.js's takeSnapshot() now calls getDecentralizationStats(), getFullDatacenterBreakdown(), getFullCountryBreakdown(), and getFullContinentBreakdown() -- each of which independently calls getAllNodeIpClassifications() (decentralizationService.js), which fully paginates the whole node_ip_classification table under Supabase. That's up to 4 full-table reads of the same data, once per snapshot cycle, in every environment.

Fix direction: one shared fetch of allClassifications + candidateIps in snapshotManager.js (or a single computeAllBreakdowns() helper in decentralizationService.js), then three in-memory group-bys instead of three separate DB round-trips.

2. Country/continent logic duplicated across 4 files

Country and continent are handled by near-identical parallel copies in:

• decentralizationService.js: getFullCountryBreakdown() / getFullContinentBreakdown() (differ only by field name)
• sqliteAdapter.js: createDecentralizationCountrySnapshots/getDecentralizationCountrySnapshotHistory and their continent twins
• supabaseAdapter.js: the same 4 functions
• snapshotManager.js: two near-identical try/catch blocks (fetch + write) for country vs. continent

Adding a third dimension (e.g. ASN, region) currently means editing all four modules, and every future bugfix has to be applied twice.

Fix direction: a single set of functions parameterized by dimension (name field, code field, table name) -- analogous to how getFullDatacenterBreakdown() already handles one dimension alone -- would remove roughly 150 duplicated lines and make future dimensions a one-file change.

Scope

• src/lib/services/decentralizationService.js
• src/lib/db/adapters/sqliteAdapter.js / supabaseAdapter.js
• src/lib/db/snapshotManager.js

Out of scope

No behavior change intended -- this is a pure refactor. Any change here needs the existing decentralization test suite (decentralizationService.test.js, nodeIpClassification.test.js, snapshotManager.test.js, decentralizationSnapshots.test.js) to keep passing unmodified in outcome, even if the internals move.

[2ndtlmining]

Problem statement

State: open, no label. Body is already a good analysis; the job is to make it executable.

Country and continent breakdowns are near-identical copies of the datacenter breakdown across four modules (decentralizationService.js, both DB adapters, snapshotManager.js), and every snapshot cycle reads the entire node_ip_classification table up to four times to serve them. Adding a third dimension (ASN, region) today means editing all four modules and applying every future bugfix twice. This is pure duplication, not a correctness bug — the numbers are right — but it is the reason the country/continent work (#138) touched so much surface area and why the reads are wasteful in every environment.

Current shape (confirmed in source)

• snapshotManager.takeSnapshot() fetches, per cycle: getDecentralizationStats(),
  getFullDatacenterBreakdown(), getFullCountryBreakdown(), getFullContinentBreakdown()
  (snapshotManager.js:182-206), then writes three snapshot sets in three try/catch blocks (:305-325).
• The three getFull*Breakdown() functions (decentralizationService.js:147-218) are near-copies that
  each call getCachedNetworkNodeIps() + getAllNodeIpClassifications() — up to 4 full-table reads of
  the same data per cycle (the 4th via getDecentralizationStats() when the cache is cold, :315-321).
• The country/continent twins in both adapters: sqliteAdapter.js:1821-1890,
  supabaseAdapter.js:1697-1790 — identical bodies with country/continent swapped.
• computeAndCacheStats() (:221-244) already contains the in-memory group-by pattern to copy.

What needs to change

• src/lib/services/decentralizationService.js:147-218 — one loadClassificationContext() (single getCachedNetworkNodeIps() + single getAllNodeIpClassifications()) and one pure computeBreakdown(relevant, dimension); keep the three existing getFull*Breakdown() exports as thin wrappers so nothing else has to change at once.
• src/lib/services/decentralizationService.js — a BREAKDOWN_DIMENSIONS map (field, codeField, sentinel, table, nameColumn, codeColumn) so a new dimension is one entry.
• src/lib/db/adapters/{sqliteAdapter.js:1781-1890,supabaseAdapter.js:1647-1790} — replace the four country/continent functions with dimension-parameterised create/read functions, resolving table and column names from a whitelist map (SQL identifiers can't be bound, so no caller input in a query string).
• src/lib/db/snapshotManager.js:182-206,305-325 — fetch the context once per cycle, derive all four breakdowns in memory, keep one try/catch per dimension around the write so a single failure still doesn't block the others.
• No migration, no API change, no chart change — pure refactor.

Implementation plan

1. One fetch, in-memory group-bys. Add loadClassificationContext() in
   decentralizationService.js returning { candidateIps, relevant } (one
   getCachedNetworkNodeIps() + one getAllNodeIpClassifications()), and a pure
   computeBreakdown(relevant, { field, codeField, sentinel }). snapshotManager calls the context
   loader once per cycle and derives all four results from it.
2. Dimension table, not parameters-in-strings. BREAKDOWN_DIMENSIONS = { datacenter: {...}, country: {...}, continent: {...} } giving { field, codeField, sentinel, table, nameColumn, codeColumn }. Keep getFullDatacenterBreakdown() / getFullCountryBreakdown() /
   getFullContinentBreakdown() as thin wrappers initially so decentralizationService.test.js and any
   other caller stay green; migrate callers in the same PR.
3. Adapters: replace the 4 functions with
   createDecentralizationDimensionSnapshots(dimension, date, rows) /
   getDecentralizationDimensionSnapshotHistory(dimension, start, end) in each adapter, resolving the
   table/columns from a whitelist map — SQL identifiers can't be bound parameters, so never
   interpolate caller input, even an internal one.
4. snapshotManager: one try/catch per dimension wrapping only the write, since the shared read is
   now single-shot; a failing write for one dimension must still not block the others (the current
   comment at :191-193 is explicit about that intent — preserve it).
5. Tables stay as they are (009_decentralization_country_continent.sql) — this is a pure refactor,
   no migration, no behaviour change. Same for the API/chart consumers.

Success criteria

• One snapshot cycle calls getAllNodeIpClassifications() exactly once (a counting test pins this; today it is up to four).
• For the same fixture, the generic breakdown returns output byte-identical to today's country/continent functions, including the (unknown) and (independent) sentinel grouping.
• The existing suites pass unmodified: decentralizationService.test.js, nodeIpClassification.test.js, snapshotManager.test.js, decentralizationSnapshots.test.js — an edit forced by the refactor would mean behaviour changed.
• Both adapters resolve table/column names through the whitelist (asserted with a mocked Supabase call), so a dimension can never inject an identifier.
• A failing write for one dimension still doesn't block the others, and a failing getAllNodeIpClassifications() still writes the snapshot row with null decentralization columns.
• Adding a fourth dimension is demonstrably a one-entry change (shown in the PR description).
• No migration, no API change, no chart change — the decentralization chart's numbers are identical before and after.

Test plan

• Regression, unmodified: decentralizationService.test.js, nodeIpClassification.test.js,
  snapshotManager.test.js, decentralizationSnapshots.test.js must pass without edits (the issue's
  own acceptance criterion). If one needs editing, that's a behaviour change — stop and reconsider.
• The redundancy half, explicitly: a test that counts getAllNodeIpClassifications() calls during
  one takeSnapshot() and asserts 1 (today's code gives 4) — this is the guard that stops the
  duplication creeping back.
• Equivalence: for a fixture classification set, the new generic breakdown produces byte-identical
  output to the old country/continent functions, including the (unknown) sentinel grouping when both
  providers omit country, and (independent) for non-datacenter orgs.
• Chained fetch failures: getAllNodeIpClassifications() throwing → snapshot row still written with
  null decentralization columns (snapshotManager.js:277-286 semantics preserved).
• Backwards compatibility: the history read functions return the same column names to the chart
  (Chart.svelte decentralization category + #138 entity search).
• Both adapters: run the adapter-level suite against SQLite for real and assert Supabase table names
  from the whitelist (mock) — a wrong table name fails silently in production.