Address potential race condition in SSHForwarder cleanup

Remove an assertion that can apparently trigger during cleanup in
rare cases, where write_eof() is called multiple times. Thanks go
to GitHub user misa-hase for reporting and helping to test this.

Author: ronf

asyncssh/forward.py

@@ -91,7 +91,8 @@ def write(self, data: bytes) -> None:
     def write_eof(self) -> None:
         """Write end of file to the transport"""
 
-        assert self._transport is not None
+        if not self._transport:
+            return # pragma: no cover
 
         try:
             self._transport.write_eof()