fix: keep recurring tasks scheduled after failed runs

[charan-rathore]

Design question

Today a recurring (interval) scheduled task keeps repeating only after a successful run. If a run fails, it is not claimed again. I could not tell whether that is intended. This PR proposes the other behavior. If success-only repetition is what you want, please close this with a word and I will not mind.

What this changes

• A failed interval task keeps its failed status and history, and becomes claimable again after the full interval. There is no immediate retry.
• The claim query now selects failed interval tasks whose next run is due.
• One-shot tasks, paused or cancelled tasks, removing the schedule, and editing the interval all stay guarded, including when a failure arrives late.
• The event copy now says the task repeats after each run, not after each successful run.

Tradeoff

A failed recurring task now keeps making provider requests at the same interval, where before it stopped. That is the cost to weigh against a task that silently stops after one bad run.

Tests

Added a test on the real SQLite store covering the due time, exclusive claim and history across two connections. It fails on the base commit. There are six scenarios in total, including the one-shot, the pause or cancel with a late failure, and removing or editing the schedule.

Results below are from the person who prepared the change; I did not rerun them. Full suite: 34 files, 162 tests pass. Format, lint, typecheck and production build pass (the build prints a chunk size warning that was already there). No dependency changes.