Fix stream() not working with base-client 0.11

[Marenz]

This PR ensures the stream() helper recreates a stream if the prior one isn’t running, and adds a test to exercise calling stream() twice.

[Copilot]

Pull Request Overview

This PR ensures the stream() helper recreates a stream if the prior one isn’t running, and adds a test to exercise calling stream() twice.

• Switches from checking broadcaster._channel.is_closed to public broadcaster.is_running in _get_stream
• Adds a second call to client.stream() in the test to catch regressions

Reviewed Changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
tests/test_client.py	Added a second client.stream(microgrid_id) call in the test
src/frequenz/client/dispatch/_client.py	Replaced protected _channel.is_closed check with is_running

Comments suppressed due to low confidence (2)

tests/test_client.py:313

• The test calls stream() a second time but doesn't assert any behavior. Consider adding an assertion that the returned stream matches the original (e.g., assert second_stream is stream) or behaves as expected to ensure idempotent behavior is verified.

_ = client.stream(microgrid_id)

src/frequenz/client/dispatch/_client.py:245

• [nitpick] The docstring for _get_stream does not mention that it will discard and recreate streams when they're not running. Consider updating it to reflect this behavior for better maintainability.

if broadcaster is not None and not broadcaster.is_running:

[llucax]

I took the liberty to change your PR description with the one Copilot created, you really need to work on your PR/commit messages, they are really not helping the reviewers at all. Not you can even ask the AI overlords to do it for you, so it shouldn't be a very complicated task 😒

[llucax]

Wouldn't it make sense to create a separate test for the double call? What if there is a bug that affects the stream() function in a way that if doesn't work when called only once but works when called twice?

[Marenz]

I split the test now