Improve HTTP MCP server observability

[MacondoExpress]

Adds structured logging for HTTP-mode diagnosis: request correlation, completion lines, middleware rejections, and tool lifecycle. Sanitizes Bolt URIs via neo4j_target while keeping raw connection keys denylisted.

Highlights:

• Server-generated request_id and per-request request completed log (status, duration, auth type, headers)
• rejectRequest helper with stable reason codes on all middleware rejections
• Tool call start/complete wrapper with duration and success flag
• Initialize verification success/failure logs; guard blocks and tool-filter debug logs
• Neo4j error neo4j_error_code on database and tool error paths
• Aura db_id inferred from Bolt hostname (*.databases.neo4j.io / neo4j-dev.io)
• SafeBoltTarget in logger; log as neo4j_target (visible), not bolt_uri (redacted)
• Startup log harmonized to listen_url; url added to denylist
• Included observability fields in log (read-only, tools filter, auth type)

[MacondoExpress]

Sorry @Liam-Doodson,
My bad, but after testing one last time, I wanted to add some others logs and I've included a proper log for the the timeout case just added, plus other bits.
Can I ask a quick review 🥺 ?
The changes:

• Added request started logs with the MCP method, making initialize, tools/list, and tools/call distinguishable.
• Changed tool-start logs from INFO to DEBUG.
• Tool failures and policy blocks now use WARNING; unexpected internal errors remain ERROR.
• Removed raw Cypher queries from logs.
• Added clear timeout warnings with the affected method, phase, duration, and tool name when available.