Pull out assay_date parsing from filename into a separate method
`parse_assay_date_from_filename`.

This is done in preparation to update the assay date parsing for VIDRL
flat files.

The assay date is the same for all measurement records from the same
fstem, so only parse the assay date once.

Use regex to find all matches for the expected date format 'YYYYMMDD'.
Then use the datetime module to validate the date string and check that
the date is earlier than the date we are parsing the file.

I made the decision to use the latest date if there are multiple matches
with the expectation that we would be parsing files not long after the
assay date.

Only fill in with the assay date parsed from the fstem when the
measurment does not have an assay date. This is done in preparation for
parsing the flat VIDRL files which will include assay date as a column.¹

This ensures that the fstem date is only a backup to the date that is
set within `format_date`.²

¹ https://bedfordlab.slack.com/archives/C03KWDET9/p1699914235686809
² https://github.com/nextstrain/fauna/blob/8088646ce0ba438310cdc9f919080950d0767c46/tdb/upload.py#L328

It's unclear if we'll ever run into the case where there are multiple
assay dates in the filename, but if we do, raise an exception to alert
the user to manually fix the filename before upload.