chore(deps): upgrade @pkcprotocol/pkc-js 0.0.45 -> 0.0.47

[Rinse12]

Upgrade @pkcprotocol/pkc-js from 0.0.45 to the latest release 0.0.47.

Plan:

1. npm install @pkcprotocol/pkc-js@0.0.47 --save-exact
2. Verify build: npm run build && npm run build:test
3. Run full test suite: npm run test:cli
4. Adapt to any API changes if tests surface them
5. Push to master

[Rinse12]

Status: upgrade to 0.0.47 installed and builds cleanly, but test/cli/daemon-kubo-restart-race.test.ts fails reproducibly with 0.0.47 and passes with 0.0.45 (A/B verified).

Diagnosis so far: with 0.0.47, kubo starts and reaches "Daemon is ready", then exits during pkc-js RPC server initialization (PKCWsServer); the daemon's keepKuboUp watchdog restarts it ~7-15s later. The test probes kubo's API right after the daemon ready banner and hits ECONNREFUSED in that window. Investigating why pkc-js 0.0.46/0.0.47 (lazy-load/perf release, pkcprotocol/pkc-js#126) causes the freshly-started kubo to exit. Not pushing until root-caused.

[Rinse12]

Root cause found. pkc-js 0.0.46 introduced new init behavior (setupKuboHttpRouters → _setHttpRouterOptionsOnKuboNode): on first connect it rewrites the kubo node's Routing config (HTTP routers) plus Provide.DHT.SweepEnabled, and when the router endpoints changed it POSTs /shutdown to kubo, logging "Clients of pkc-js should restart ipfs node".

On a fresh repo the endpoints always differ from kubo defaults, so kubo is shut down ~50ms after reaching "Daemon is ready". The daemon's keepKuboUp machinery restarts it correctly (this is the contract pkc-js expects), but daemon-kubo-restart-race.test.ts sets PKC_CLI_TEST_KEEPKUBOUP_PORTCHECK_DELAY_MS=7000, stretching that restart to ~7s — and then probed kubo's API immediately after the daemon's ready banner, hitting ECONNREFUSED in the down-window. Verified by capturing the daemon log: kubo receives the shutdown 51ms after startup, exactly during PKCWsServer init.

Fix: the test now waits (waitForKuboReady, 40s cap) for kubo to survive the pkc-js-induced restart before killing it to trigger the race under test. The daemon code itself needs no change. Full suite re-running with 0.0.47.

[Rinse12]

Done: upgrade committed and pushed to master as 555cdde. Full suite green (38 files, 293 passed, 1 skipped). The only adaptation needed was in daemon-kubo-restart-race.test.ts, which now waits for kubo to survive the pkc-js-induced router-reconfig restart before triggering the race under test.

[Rinse12]

CI fallout from the upgrade fixed in 1849ff8: the macOS failure (test's kubo /shutdown colliding with pkc-js's own router-reconfig shutdown — "other side closed") is absorbed by a tolerant requestKuboShutdown helper used at all five bare /shutdown call sites, and the Windows failure (community get observing pre-edit state right after community edit) by polling reads in the edit null-removal e2e until the post-edit state is visible. The pre-existing ubuntu failure was a separate bug, tracked and fixed in #92. Watching the CI run on the new commits.

[Rinse12]

CI run 27414930318 on 28591bd is green on ubuntu, macOS, and Windows — the upgrade plus test hardening is fully verified.