Merge retries and testthat

[wlandau]

This PR has:

1. Better handling of failed merges such as Contribute remotes contributions#145. If a merge fails, the error message is reported, and the PR is not flagged for manual review. That way the bot can just retry next time. I do not think manual review is necessarily required if the bot already approved the PR, but it will need to do another automated review in case the contributor made any changes between rounds.
2. Uses testthat. It's only a soft dependency now, and it's a much more ergonomic developer experience in my opinion.
3. Removes some duplicate tests I found. Not sure why/how those got copied.

[shikokuchuo]

I'm fine with the concept here of re-trying the next time.

(i) Could there be a problem where it keeps re-trying every hour? Potentially no one is monitoring the bot that often. Does it send an email to the submitter every time? That could become problematic.

(ii) do we know all the potential error messages that could be displayed? We don't want the possibility of leaking secrets or other internal information.

[wlandau]

Great points on both. I just added a commit that omits the error message and adds a special retry-review label when a merge fails. Issues labeled retry-review should not get additional messages about merge failures.

[shikokuchuo]

Ok sounds good.