feat(storage): add persistence for schedule_runs, alert_events, and candidate_suggestions

[CrisSTEM]

Summary

Add the SQLite migration and storage-layer persistence helpers for schedule_runs, alert_events, and candidate_suggestions.

This PR covers fresh bootstrap, upgrade from the current ingestion baseline, representative persistence operations, and preservation of existing observation history across SyncPack runs.

Linked issue(s)

• Closes feat(storage): add persistence for schedule_runs, alert_events, and candidate_suggestions #46

Milestone fit

This PR adds the storage substrate required for the v0.4.0 backend-processing baseline while keeping scheduler orchestration, alert evaluation, candidate discovery heuristics, and API/frontend read models intentionally out of scope.

It also preserves the current separation between catalog/config synchronization and persisted observation history.

What changed

• added SQLite migration 003_processing.sql for:
  • schedule_runs
  • alert_events
  • candidate_suggestions
• added storage helpers for:
  • schedule run lifecycle persistence
  • alert event insert/resolve flows
  • candidate suggestion upsert/status update flows
• added deterministic tests for:
  • fresh bootstrap
  • upgrade from the v0.3.0 schema baseline
  • representative persistence operations
  • preserving persisted history across SyncPack catalog updates

Out of scope

• scheduler execution and due-check orchestration
• alert rule evaluation logic
• candidate discovery heuristics
• API or frontend read models
• new source adapters

Validation

Mark what you actually ran:

• python3 ./scripts/trim-whitespace.py --check
• go vet ./...
• go test ./...
• make ci
• Manual CLI verification was performed if runtime behavior changed

Documentation impact

• No docs update needed
• README.md
• SPEC.md
• docs/
• repository governance docs (AGENTS.md, CONTRIBUTING.md)
• CHANGELOG.md
• release notes

Reviewer focus

Call out the highest-risk files, behaviors, or assumptions to review.

• migrations/003_processing.sql for constraints, foreign keys, and index shape
• internal/storage/processing.go for writer/read helper boundaries and config-context lookups
• internal/storage/processing_test.go for upgrade-path coverage and preservation of persisted history across SyncPack

Auditability checklist

• Branch name follows project convention
• PR scope matches the linked issue and milestone
• Commits were pushed progressively, not as one final dump
• Relevant tests were added or updated for behavior changes
• No secrets or sensitive credentials were introduced
• Deferred follow-up work is documented explicitly
• Unimplemented work is not implied as completed

Notes for merge

Preferred merge method:

• Issue PRs: Rebase and merge
• Release aggregation PRs: Merge commit
• Use Squash and merge only when the branch became noisy and preserving intermediate commits no longer adds value