feat: add Store abstraction for build data access

[Scr4tch587]

What

Put the build path's four data ops behind a Store ABC with two backends, and thread a store through the orchestrator and worker.

How

• Store ABC over get_existing_timestamps, get_rows_range, get_rows_timestamps, insert_rows + build_lock
• PostgresStore — thin shell forwarding to core.db.datasets
• MemoryStore — in-process dict backend; json round-trip on insert mirrors Postgres Jsonb
• store threaded run_build → execute_job → _fetch_dep_data
• worker no longer calls core.db.datasets directly
• store defaults to PostgresStore(), so existing callers are unchanged

Why

Lets the same build logic run against Postgres or an in-memory store, the groundwork for dry runs. No behavior change on its own.

[github-actions]

• main
  • feat: add Store abstraction for build data access 👈
    • feat: add dry-run flag to build_dataset #353
      • feat: add dry-run query param to build endpoint #354
        • docs: document dry-run builds in backend spec #355