fix(stream): close the file handle when save_as fails

[Mohammed Alkindi (MohammedAlkindi)]

Stream.save_as closes the file only after the read loop finishes, so an exception from self._channel.send("read", ...) skips the close. Connection.cleanup() raising TargetClosedError into a pending callback is the ordinary way that happens.

The sibling handles the upload direction properly: WritableStream.copy wraps its loop in with open(path, "rb") as f:.

The impact is narrower than it first looks, so to be precise: on CPython an uncaught exception drops the frame and the handle closes anyway. The difference shows when the caller keeps the exception, since __traceback__ retains the frame and its locals:

current       file.closed == False
try/finally   file.closed == True

That is the realistic shape — except Error as e: holding e keeps the frame — and on Windows an open handle also blocks removing or reopening the partial file.

black --check, flake8 and isort --check pass on the changed file, at the versions pinned in .pre-commit-config.yaml.

Could not verify: I did not run the repo's pytest suite; it needs the Node driver, which this machine cannot fetch. The measurement above reproduces the same control flow standalone.