Use /openapi.json for mem0 healthcheck instead of /memories

[JartanFTW]

Problem

The mem0 backend healthcheck calls GET /memories?user_id=__health_check__ every 30 seconds (Docker/deployer healthcheck interval). This shows up as constant polling in mem0's request activity logs, creating noise that obscures real traffic.

Current Workaround

None. Deployer could increase the healthcheck interval, but that delays outage detection.

Proposed Solution / Alternatives

Switch the healthcheck target from GET /memories to GET /openapi.json.

• Pros: Read-only, no side effects, no mem0 data plane noise, lightweight
• Cons: Only proves mem0's HTTP layer is up, not the database/vector store behind it
• Tradeoff accepted: Shallow-but-quiet is the right call for a 30s interval. Actual backend failures surface on real requests anyway.

Change in memcp/backend/mem0.py health() method — single line swap.