fix(core): remove double offset in GetHistoricalServerStats pagination

[CaiJingLong]

Related Issue

Closes #1532

Problem

GetHistoricalServerStats applies the page offset twice: once when building ts_vec (shifting the time window back by granularity * STATS_PER_PAGE * page), and again via .skip(page * STATS_PER_PAGE) on the MongoDB query. Since the { ts: { $in: ts_vec } } filter already restricts the result to at most STATS_PER_PAGE documents for that page, the subsequent .skip(page * STATS_PER_PAGE) discards every matched document when page >= 1, so every page after the first always returns an empty stats array and next_page becomes None, silently losing all older historical data.

Change

• bin/core/src/api/read/server.rs, GetHistoricalServerStats::resolve: removed the .skip(page * STATS_PER_PAGE) call from FindOptions::builder(). ts_vec already encodes the page offset via the time window shift, so the $in filter is the correct paging mechanism here — no additional skip is needed.

Verification

• cargo check -p komodo_core passes (no new warnings).
• cargo fmt -p komodo_core --check passes.
• Cross-checked against the issue reproduction: page = 0 behavior is unchanged (.skip(0) was already a no-op); page >= 1 is no longer wiped out by the second offset and now returns the documents matching the ts_vec time window, with next_page correctly indicating whether an older window has data.
• bin/core has no existing unit-test infrastructure and the function depends on a live MongoDB runtime, so no new tests were added; the fix can be verified manually via the issue's 1-min multi-page reproduction steps.

Risk

• Change is a single-line removal; does not affect the page = 0 path. The ts_vec construction and next_page logic are unchanged. No schema migration, config change, or breaking API change.
• Note: next_page is Some(page + 1) only when the page returns exactly STATS_PER_PAGE records. If a page returns fewer than 200 due to missing data while an older window still has data, next_page will be None — this is pre-existing behavior, unchanged by this fix, which only removes the double offset that caused the spurious empty results.

Conventions

• The repository has no explicit CONTRIBUTING.md / CLAUDE.md / AGENTS.md / PR template, so common conventions (Conventional Commits + default PR structure) are used.