feat(ops): configure litellm prompt logging

[casey-brooks]

Summary

• add LiteLLM proxy_config.yaml with prompt/model storage defaults
• mount the config into the litellm service and start it with --config
• drop the redundant STORE_MODEL_IN_DB env so YAML is the source of truth

Testing

• docker compose up -d litellm-db litellm
• curl -sS http://127.0.0.1:4000/chat/completions -H 'Authorization: Bearer sk-dev-master-1234' -H 'Content-Type: application/json' -d '{"model":"gpt-4o-mini","messages":[{"role":"user","content":"log-test: say hello"}]}'
• curl -sS http://127.0.0.1:4000/spend/logs -H 'Authorization: Bearer sk-dev-master-1234' (shows request content)
• /nix/var/nix/profiles/per-user/root/profile/bin/pnpm lint (fails: existing @typescript-eslint/no-unsafe-assignment errors in packages/platform-server/src/infra/container/runnerGrpc.client.ts)

Local LiteLLM verification: ✅ request/response now appears in spend logs (mirrors UI Logs)

Resolves #1360

[casey-brooks]

Test & Lint Summary

• docker compose up -d litellm-db litellm
• curl -sS http://127.0.0.1:4000/chat/completions -H 'Authorization: Bearer sk-dev-master-1234' -H 'Content-Type: application/json' -d '{"model":"gpt-4o-mini","messages":[{"role":"user","content":"log-test: say hello"}]}'
• curl -sS http://127.0.0.1:4000/spend/logs -H 'Authorization: Bearer sk-dev-master-1234' (shows request content)
• /nix/var/nix/profiles/per-user/root/profile/bin/pnpm lint (fails: existing @typescript-eslint/no-unsafe-assignment errors in packages/platform-server/src/infra/container/runnerGrpc.client.ts)

[noa-lucent]

Looks good. This matches the linked issue: config is now provided via proxy_config.yaml, mounted into the LiteLLM container, and loaded with --config; prompt/model storage defaults are defined in YAML and the redundant STORE_MODEL_IN_DB env var was removed.