[OMEGA-503] feat(rag): add opt-in background knowledge indexing

[amxson]

Description

Knowledge indexing currently blocks startup until changed files are embedded. This adds opt-in knowledgeIndexingBackground so plugins and channels can start while knowledge becomes available progressively.

Indexing uses the runtime Chroma collection. The setting defaults to false, preserving the existing synchronous behavior.

Closes #384.

How Has This Been Tested?

• Added coverage for synchronous/background indexing, duplicate workers, and failure logging.
• Verified progressive retrieval and runtime memory with persistent Chroma.
• Python, PeTTa, and mandatory autotests pass.

Checklist

• PR contains autogenerated code
• Self-review completed
• Test scenarios above are passed with the version of the code from PR

[timur-ashkenov]

@amxson Is dropping CHROMA_DB_PATH here intentional? Import/export still use it, while lib_chromadb uses ./chroma_db. Could that make them point to different databases when a custom path is set?

[amxson]

Good catch. I restored CHROMA_DB_PATH handling while keeping RAG and runtime on the same shared collection. I also resolve the paths before comparing them so equivalent paths don’t open another client.

[timur-ashkenov]

The thread drops the result returned by _index_knowledge(), so we never see the final summary or “no files found” message. Could we log it when the worker finishes?

[amxson]

Yep, fixed. The background worker now logs the result from _index_knowledge(), including the normal summary and the no-files case.

[timur-ashkenov]

@amxson Both fixes look good, thanks! Could you update the branch with the latest main so we can run CI before merging?

[amxson]

Updated with latest main. Local tests pass, including the mandatory suite.