Consider how to handle API request failures in `fetch_ci_data.py`

[liz-is]

Depending on the type/frequency of errors we observe in practice, it might be easier to add rate limiting and retry logic here, rather than rerunning the cron job.

I’ve also recently heard about tenacity; but that’s probably more than we need here.

Originally posted by @JostMigenda in #11 (comment)