Restore the scraper public boundary and durable architecture docs

Problem

The scraper design calls for one public lifecycle boundary, but routes, schemas, serializers, and tests import configured/service, configured/rules, configured/runtime, and other internals directly. Durable architecture also remains mainly in completed OpenSpec change documents rather than docs/architecture and the module README.

Relevant code:

• apps/server/src/scraper/index.ts
• apps/server/src/scraper/README.md
• apps/server/src/orpc/routes/external-sites/scrape-sources/
• apps/server/src/schemas/scrapeSources.ts
• apps/server/src/serializers/scrapeSource.ts
• openspec/changes/define-scraper-runtime/
• openspec/changes/add-configured-scraper-sources/

Desired outcome

Callers depend on small supported scraper APIs, while implementation modules can move or be deleted without application-wide import changes. Durable docs describe the architecture that exists today.

Acceptance criteria

• Define explicit runtime and management entry points with only the operations/types used by outside code.
• Move production callers off deep imports into scraper/configured/*.
• Add a lightweight boundary check that prevents new deep production imports; test-only imports may be explicitly allowed.
• Promote current runtime, queue, ownership, source/target/origin, versioning, and test boundaries into docs/architecture and link from the scraper README.
• Document that scraper jobs use the dedicated scrapers queue but currently share a process with the default worker.

Non-goals

• Create a broad barrel that exports every helper.
• Split the worker into another service without measured resource-isolation need.

Start with apps/server/src/scraper/index.ts and README.md, then trace the production imports under the listed routes, schemas, and serializers. Review the two OpenSpec change directories for the current runtime and ownership boundaries. Done means supported runtime and management entry points are used by production callers, a boundary check exists, and durable architecture documentation is linked from the scraper README.