Clean up unused schema fields and indexes

[mrubens]

Summary

• drop unused task_runs.base_shas, deprecated task_messages.content_schema, and unwritten webhooks.ignored_at
• simplify the webhook status exclusivity constraint to the supported ok / error states
• remove nine non-unique indexes that exactly duplicate retained unique indexes or constraints
• update fixtures and assertions that referenced the removed columns

Migration verification

Applied the full migration chain to a disposable Postgres database and verified:

• all three unused columns are absent
• all nine duplicate indexes are absent
• all nine unique backing indexes remain
• no exact duplicate indexes remain in the generated schema snapshot

Validation

• DB tests: 218 passed
• webhook tests: 8 passed
• task-title refresh tests: 5 passed
• affected package typechecks and lints
• pnpm lint:fast
• pnpm check-types:fast
• pnpm knip

[roomote-roomote]

No code issues found. See task

Reviewed the schema cleanup against the codebase: the three dropped columns (task_runs.base_shas, task_messages.content_schema, webhooks.ignored_at) have no remaining references in application code, and the migration ordering (drop constraint → drop columns → re-add webhooks_status_exclusive) is correct. Eight of the nine dropped indexes are exact duplicates of retained unique constraints/inline .unique() columns; the ninth (repositories_provider_full_name_idx on (source_control_provider, full_name)) is not an exact unique duplicate, but its query workload (findRepository) is already covered by the more-selective repositories_full_name_idx, so its removal is safe.