Bucket-backed DB persistence + dual-workspace Phase 1 (Workspace, Stage, Outreach)

[futureaigh]

What

Storage: Bucket, not Volume. The SQLite DB file (Prisma stays — it has no DuckDB provider, so an engine swap would mean rewriting 100+ call sites for zero single-user benefit) now persists to a Railway Bucket (jobsync-db, sjc) instead of the Railway Volume:

• scripts/sync-db-bucket.mjs: dep-free SigV4 S3 client (GET/PUT/HEAD), download on boot, upload on 5-min timer + SIGTERM/SIGINT, syncs main file + WAL sidecars, exits 0 when unconfigured (local dev unaffected)
• docker-entrypoint.sh + Dockerfile: restore before prisma migrate deploy, periodic push, scripts/ copied into the slim runner (zero new npm deps)
• Service vars set via Railway CLI: AWS_ENDPOINT_URL, AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY, AWS_S3_BUCKET_NAME, AWS_DEFAULT_REGION=auto, S3_PATH_STYLE=false

Dual-workspace Phase 1 (PRD): Workspace/Stage/WorkspaceField/Outreach models + migration dual_workspace_outreach; workspace server actions (create with default SCHOOL/JOB pipelines, idempotent ensure-defaults, guarded delete); outreach actions (logOutreach auto-sets follow-up = sent + 12d, getFollowUpsDue, markOutreachReplied); followUpDueDate helper encoding the 10–14d rule; osui WorkspaceSwitcher component; seed-school-workspace.mjs seeding the SCHOOL workspace + 7 PI contacts (verified idempotent against local dev DB).

Tests

• 27 new tests, all passing: bucket-sync (17, mocked fetch incl. SigV4 shape/determinism), outreach-follow-up, workspace.actions, outreach.actions
• tsc --noEmit clean; prisma validate + migrate dev + generate green

Follow-ups (not in this PR)

• Detach jobsync-volume after first successful bucket-backed deploy
• Phases 2–5: workspace routing/dashboard, applications board/table/calendar, PI directory UI, MCP tools