Fix concurrent socket connection issue

[jaredcnance]

Closes #40

This makes the TCP client safe to use across threads by allowing only a single thread to connect at a time. Additionally, it adds a single connect+retry for messages that fail because of a disconnected peer. Note that there is currently an issue with the test_can_recover_from_agent_shutdown that is a result of the test socket not being properly cleaned between shutdown and resume. This causes the test agent to fail to start occasionally.

[hussam789]

PR Code Suggestions ✨

Category	Suggestion	Impact
Possible issue	Fix potential infinite recursion The retry counter is initialized to 1 but never decremented in the initial connection attempt. If the first connection fails, you'll still have retry=1 when recursively calling send_message, potentially leading to infinite recursion if connection keeps failing. aws_embedded_metrics/sinks/tcp_client.py [61-67] def send_message(self, message: bytes, retry: int = 1) -> None: if retry < 0: log.error("Max retries exhausted, dropping message") return if self._sock is None or self._sock._closed or self._should_connect: # type: ignore self.connect() + if self._should_connect: # If connection failed + return self.send_message(message, retry - 1) Apply this suggestion Suggestion importance[1-10]: 8 __ Why: The suggestion identifies a potential infinite recursion bug by not decrementing the retry counter when the initial connection fails and proposes a change that handles this case correctly.	Medium
More