Have it run for the target branch with a command

[joao-paulo-parity]

If a developer wants to compare how the current version of the bot is behaving against the target branch e.g. master, there's no way to do that with the current commands: they would need to open a useless pull request just for that.

There should be a command in the likes of /bench target-branch ... which runs the benchmarks against the target branch and posts the weights diff as a comment.