fix: harden Synctech Drive sync lifecycle

[danshapiro]

What changed

• Scheduled Synctech SMS Drive syncs now use the daemon's already-open store when run from serve.
• Drive polls create one outer source-level sync_runs row, including zero-selected polls, and fail that run on list/checksum/stability/download/import/checkpoint errors.
• Drive imports now import backup files into the existing source lifecycle instead of creating nested sync runs per downloaded backup.
• Successful Drive polls update checkpoint counts, touch sources.last_sync_at, and recompute conversation stats when files import.
• Scheduled cache rebuild logic is shared by Gmail/IMAP and configured Synctech syncs.

Why

This keeps scheduled Drive sync lifecycle state accurate and avoids duplicate stores/runs that can obscure failures or freshness.

Impact

Configured Drive sync status, freshness, and analytics cache updates should now reflect every poll, including no-op polls and failed polls.

Validation

• go test -tags "fts5 sqlite_vec" ./cmd/msgvault/cmd ./internal/synctechsms ./internal/store
• make testify-helper-check
• go vet -tags "fts5 sqlite_vec" ./...

[roborev-ci]

roborev: Combined Review (2129ac9)

Summary verdict: No Medium, High, or Critical findings to report.

The only reported issue was Low severity and is omitted per the review-combination rules. Security review found no issues.

---

Panel: ci_default_security | Synthesis: codex, 7s | Members: codex_default (codex/default, done, 3m57s), codex_security (codex/security, done, 2m36s) | Total: 6m40s