fix(sync): cascade series exceptions on provider delete

[tyler-dane]

Summary

• After Google confirms a provider series delete (scope=all), Sync now cascades findSeriesExceptions: clear occurrence rows, delete exception events, then delete the master (same order as cloud series delete).
• Already-gone replay path also clears leftover exceptions so older staging orphans and crash mid-cascade residue heal on retry.
• DB coverage: series master + exception → both gone; orphan exception with missing master → cleared on replay.

Why

Staging cutover found Google instance overrides left active after Compass deleted the series. Pull cannot heal once the master is gone. Browser DELETE 503 with backend 204 remains reporting noise; this is the real blocking Sync gap.

Test plan

• bun test:sync -- packages/sync/src/domain/provider-command.service.db.test.ts
• bun test:sync (full suite)
• Staging re-gate after deploy: create weekly → Google edit one instance → delete scope=all → range has neither master nor exception; orphan scan clean; CRUD smoke

Made with Cursor