fix(signal): close httpx client on health check failure to prevent connection leak

[binhnt92]

When connect() creates an httpx.AsyncClient and the subsequent health check fails — either a non-200 response or a network error — both return False paths exit without closing the client. The TCP connection pool, SSL context, and file descriptors are leaked. On repeated reconnect attempts (e.g. the gateway retrying after a daemon restart), this accumulates open connections that are never cleaned up.

Changes Made

Added await self.client.aclose() and self.client = None on both failure paths in the health check block (gateway/platforms/signal.py).

How to Test

python3 -m pytest tests/gateway/test_signal_client_leak.py -v

4 tests: client closed on non-200, client closed on connection error, client kept on success, source verification.

Checklist

• Tests added (4 tests)
• Full test suite run — no regressions
• Tested on Linux (Ubuntu 22.04)

[dieutx]

I'm working on Signal resource management in #3692 (FIFO eviction for sent-timestamp tracking). The httpx client teardown here looks solid — just want to note that if the health check fires while a message is being sent and tracked in _recent_sent_timestamps, closing the client mid-flight could cause the send to fail silently, and the timestamp would still be added to the tracking set. Might be worth checking if there's an in-flight send before closing.

[binhnt92]

Looks superseded by #5519 now. Ok to close this one.