BrokerClient: add detailed ensure_connection errors and timeouts

[oboehmer]

Summary

BrokerClient.ensure_connection() currently collapses all broker/connect/protocol failures into False, with only debug/error logs depending on where the failure occurs. This makes broker-related failures harder to diagnose, and there are no explicit read/write timeouts around socket operations (risk of indefinite hangs in worst-case failure modes).

This is low priority (broker is best-effort / fallback exists), but it would improve debuggability and robustness.

Current Behavior

• ensure_connection(hostname) returns False on any error (no reason provided to caller).
• Failures are logged at debug in ensure_connection(), and at error in connect() for some connection failures.
• No explicit timeouts around open_unix_connection() / StreamReader.readexactly().

Proposed (Low-prio) Improvements

Option A (preferred):

• Keep ensure_connection() -> bool for compatibility
• Add ensure_connection_detailed() -> tuple[bool, str] returning a reason string on failure
• Add socket timeouts (configurable; default small) and tests proving no indefinite hang

Option B:

• Return (bool, error_msg) from ensure_connection() directly (API change)

Test Ideas

• Stale socket file exists but no server listening
• Broker process dies mid-test (connection reset)
• Partial response / protocol error
• Timeout on response length / response body read

Related

• [ConnectionBroker] Add failure mode test coverage #486 (failure mode test coverage)