Update unknown_facilities list generation to include filename

[cawarren]

Issue
The script currently (and incredibly helpfully) outputs a list of unrecognized facility names after each merge.

Many of these names just need to be added to the facility names mapping sheet, but some are indicative of a collection problem. For example, in the screenshot below it's clear that a scraper has gotten its columns mixed up:

Unfortunately we currently have no way to tell which file from the covidprisondata.com dataset that mix-up occurred in, so we can't easily point our collaborators towards the issue.

Request
If easy (this is P2), please update the aggregator to include a fourth column in the unknown_facilities.csv output: the filename the facility was found in.