[Refactor] Split local model discovery out of task-models settings

[roomote-roomote]

Created by Roomote. Follow up by mentioning @roomote-roomote or in the web UI.

Related issue

Follow-up to the self-hosted inference work in #481: local endpoint discovery had grown into the task-models settings command monofile.

Why this PR exists

• A maintainer explicitly invited this PR in the linked issue or discussion
• I am a maintainer / this is internal Roomote work

What changed

Extracts Ollama, vLLM, and LiteLLM model discovery, recommendation scoring, and qualify probes out of the task-models settings command module into a dedicated local-provider discovery helper.

tRPC command entry points stay as thin admin wrappers with the same public APIs. Endpoint provider ids and connection env var names now come from the shared setup provider catalog instead of a second hard-coded list.

Matching discovery, scoring, and qualification tests moved with the extraction. Behavior should be unchanged for operators connecting local/endpoint providers.

How it was tested

• Targeted web Vitest for local-provider discovery and task-models commands (46 passed)
• Web package typecheck
• Pre-push gates: oxlint, residual lint, check-types:fast, knip

Impact

No intended user-facing behavior change. Maintainers get a smaller settings command module and a clear place to extend local/endpoint discovery without mixing it with settings CRUD.

Checklist

• The PR title follows the repo convention: [Fix], [Feat], [Improve], [Refactor], [Docs], or [Chore] followed by a user-facing description
• This PR is small and scoped to one change
• pnpm lint and pnpm check-types pass locally
• I added tests or included a clear manual validation note above
• I removed secrets, tokens, private keys, and customer data from code, logs, and screenshots
• If this change should appear in the changelog, I ran pnpm changeset

[roomote-roomote]

No code issues found. See task

_{Reviewed 17f5f44}