Store-object convergence re-runs hourly on a plain-PostgreSQL store too, not only at restart (#3913)

[erikdarlingdata]

Closes #3913.

CHANGELOG: this PR carries no CHANGELOG edit. Its entry lands with the batched splice.

Why

#3817 made store-object convergence re-run hourly, so a failed ensure heals within the hour. That hourly pass sat inside the tick's _timescaleAvailable gate, so on a store without TimescaleDB, a supported configuration, nothing re-ran between restarts. The affected steps are:

• a dropped baseline fallback view, which leaves one anomaly family silently returning nothing;
• a retired baseline relation's time-dependent drop;
• the store's statement statistics (Field stores cannot tell us which query is slow: preload pg_stat_statements in the managed store and surface it by role #3899): an extension a DBA creates by hand, or a grant for a role provisioned later;
• the composer's covering indexes.

Each waited for the next restart.

What changes

• ConvergeStoreObjectsAsync takes timescaleAvailable and has one filter. Without TimescaleDB it skips exactly the stages the start path runs inside its own TimescaleDB gate (NeedsTimescale: Timescale, TimescaleAfterRepairLaunch). Every other step runs, as it does at start.
• The tick gains an else branch that runs the pass on a plain-PostgreSQL store. The gated block, its four-tenant order and the #3575 phase argument are untouched; none of that applies to a store with no policy jobs.
• The cadence text Field stores cannot tell us which query is slow: preload pg_stat_statements in the managed store and surface it by role #3899 added ("a plain-PostgreSQL store waits for its next start") is corrected in get_store_query_stats' remedies, the step comment, StoreStatementStats, provision-roles.sql and the README.

Test plan

• The convergence pin now asserts the hourly pass's ONE filter and no other stage test; NeedsTimescale is pinned per stage.
• The tick pins (TimescaleAvailabilityReprobeTests, RetentionReevaluationTests) are unchanged and green: one stamp, one gated convergence call, same order.
• Full Darling.Tests after rebasing onto The store's statement text and store log keep no literal: a one-pass SQL lexer, an allowlist reader, masking for every retained log class, and stricter conf parsing (#3899) #3920: 12,487 run. The only failures are the three runtime-version pins my local rig (18.4) can't meet against dev's 18.6 pin; CI builds the pinned runtime.

🤖 Generated with Claude Code

https://claude.ai/code/session_018Z4T51J4E6L8fjriPHpsif