Fix UDS failing if socket already exists

In trying to use UDS in the Datadog injector test suite, I discovered
that the UDS socket functionality doesn't work particularly well in
containers. I suspect this is due to the atexit hook not being called
due to the abrupt container shutdown process. Subsequent runs of the
container will then fail as the file already exists and results in a
`OSError: [Errno 98] Address already in use` exception.

ddapm_test_agent/agent.py

@@ -1263,9 +1263,10 @@ def main(args: Optional[List[str]] = None) -> None:
 
     apm_sock: Optional[socket.socket] = None
     if parsed_args.trace_uds_socket is not None:
+        if os.path.exists(parsed_args.trace_uds_socket):
+            os.unlink(parsed_args.trace_uds_socket)
         apm_sock = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM)
         apm_sock.bind(parsed_args.trace_uds_socket)
-        os.chmod(parsed_args.trace_uds_socket, 0o722)
         atexit.register(lambda: os.unlink(parsed_args.trace_uds_socket))
 
     if parsed_args.trace_request_delay is not None:

releasenotes/notes/uds-c99ea5d099e67156.yaml

@@ -0,0 +1,5 @@
+---
+fixes:
+  - |
+    Remove any existing UDS socket. If a socket was left behind by a previous
+    instance of the test agent then it would fail to start.