fix(meerkat-dbm): stop idle shutdown from terminating the worker mid-query

[shriram-devrev]

What

Two layered changes so an idle timer can't terminate the duckdb-wasm worker while a query is running:

1. Cancel the pending recycle/shutdown timers when a query starts (_startQueryQueue). They re-arm on the next queue drain. Primary fix — no idle timer is left pending during a query.
2. Guard the shutdown timer callback on _isBusy() (queue length OR queue running OR currentQueryItem), matching the recycle timer. Covers the event-loop edge where the timer callback was already queued before the new query cleared it.

Also removes the dead _isStaleWorkerError retry from #287.

Why (root cause, verified against prod)

The shutdownInactiveTime timer is armed only when the queue drains (_stopQueryQueue), and was never cleared when the next query started. So a timer armed on the previous drain kept counting down through the start of the next query. _startQueryExecution shifts the query off the queue before running it, so during execution queriesQueue.length === 0 while currentQueryItem is set and queryQueueRunning is true. The timer only checked queriesQueue.length > 0, so it treated the in-flight query as idle and called _shutdown() → terminateDB() mid-query.

duckdb-wasm does not throw for this — AsyncDuckDB.postTask on a detached worker only console.errors "cannot send a message since the worker is not set!" and resolves. Confirmed in prod RUM: every event is source: console, handling-stack console error → postTask. That's also why #287's catch-based self-heal never fired — the 0.1.45 build was empirically a no-op (fixed vs unfixed builds showed identical error rates).

Tests

• does not terminate the worker when the shutdown timer elapses mid-query — query held in flight (slow preQuery) past shutdownInactiveTime must not trigger terminateDB, then shuts down normally once idle.
• cancels the pending shutdown timer when a new query starts — a second query started before a prior armed timer elapses must clear it so it never fires.

Full meerkat-dbm suite green (20 dbm.spec + all others); nx build + nx lint clean.

Notes

• Bumps @devrev/meerkat-dbm 0.1.45 → 0.1.46. devrev-web bumps to pick it up.
• Scope: single-DBM race (the dominant production case). The cross-DBM shared-engine variant (multiple DBM types swapping one field) is out of scope here.
• Supersedes fix(meerkat-dbm): self-heal stale DuckDB worker in query() #287 (no-op) and fix(meerkat-dbm): detect stale worker by engine identity, not error message #289 (isDetached — reconnects after the fact but doesn't prevent the mid-flight teardown). Both closed.

work-item: ISS-334477