Bump knowrag-ollama image to 0.30.0-rc20 for qwen3-embedding support

[w7-mgfcode]

Summary

The ollama service in @lab/ll-KNOWRAG/compose.yml is pinned to ollama/ollama:0.5.4. That release predates the qwen3-embedding model family — pulling qwen3-embedding:4b fails with:

Error: pull model manifest: 412: The model you are attempting to pull requires a newer version of Ollama.

Change

• Bump the pinned image: ollama/ollama:0.5.4 → ollama/ollama:0.30.0-rc20.
• Update the Current pins list in @lab/ll-KNOWRAG/.claude/rules/docker-infra.md to match.

Acceptance criteria

• docker compose --env-file .env.example config -q validates.
• docker exec knowrag-ollama nvidia-smi still reports the RTX 5060 (GPU reservation intact).
• docker exec knowrag-ollama ollama pull qwen3-embedding:4b succeeds.
• Existing models (nomic-embed-text, qwen2.5:*) remain listed.

Zone

@lab — disposable sandbox. 0.30.0-rc20 is a release-candidate, chosen deliberately for the latest model support.