docs: update backend spec for scheduler/worker/orchestrator architecture

[Blackgaurd]

Updates SPEC-backend.md to document the new scheduler/worker/orchestrator build architecture.

What changed

• New "Build architecture: scheduler / worker / orchestrator" section documenting:
  • Scheduler: DFS graph collection + Kahn's algorithm for topological ordering
  • Worker: single-job execution with per-dataset locking and atomicity
  • Orchestrator: level-by-level plan execution with barrier model
  • Data types: `JobDescriptor`, `JobResult`, `BuildPlan`
  • Commit-per-level model
• Updated directory layout to include new service modules (`scheduler.py`, `worker.py`, `orchestrator.py`, `models.py`, `timestamps.py`, `locks.py`)
• Updated atomicity section: per-job atomicity with per-level commit
• Updated build concurrency section: lock scope now in `execute_job()`, deadlock safety guaranteed by startup DAG validation + topological execution order
• Updated logging section: split across scheduler, orchestrator, worker
• Updated calendar integration: `generate_timestamps()` now in `service/timestamps.py`

🤖 Generated with Claude Code

[Blackgaurd]

Warning

This pull request is not mergeable via GitHub because a downstack PR is open. Once all requirements are satisfied, merge this PR as a stack on Graphite.
Learn more

• docs: update backend spec for scheduler/worker/orchestrator architecture #324 👈 (View in Graphite)
• refactor: replace _build_recursive with orchestrator #323 : 1 other dependent PR (#325 )
• feat: add build orchestrator #322
• feat: add build worker that executes a single job #321
• feat: add topological scheduler with Kahn's algorithm #320
• feat: add dependency graph collector with range expansion #319
• refactor: add job descriptor and build plan types #318
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.