Our unit tests dump ~536 lines to stdout because bots print their logs. This verbosity makes it difficult to glean important information from our test output.

This adds a bot_log contextmanager to capture bot logs and let us assert about their content. This has a couple advantages...

1. Less verbose test output.
2. Better test coverage by ensuring our bots emit the right output.

If you like this direction I'll do the same for our other tests, but first I need to confirm that the project owner approves of this.