Rhino caches API rate limite responses

[GuySartorelli]

Found when running the merge-ups task, but likely happens with any of them.

If you forget to add the github token in your .env file (or for any other reason you hit rate limits), the rate limit response is cached in the database. Rerunning the task after adding the token (or waiting for the rate limit timeout) fails, because Rhino fetches the cached response and packs a sad because it thinks you're still hitting the rate limit.

I also didn't get a clear error message when running the task - I had to debug to find out why I was getting server errors suddenly.

Acceptance criteria

• Failed API responses (including but not limited to rate limiting) are not cached in the database
• Failed API requests result in a clear message being displayed in the browser (for tasks) or in the job message list (for queued jobs)