CivicPlusSite needs better handling for names of downloaded files

[zstumgoren]

On a test scrape for Belvedere, CA for roughly June through early August, the scrape generated less-than-helpful names for downloaded files:

/tmp/civic_scraper/
├── assets
│   ├── civicplus_www_06142021-577_agenda.html
│   ├── civicplus_www_06142021-577_agenda.pdf
│   ├── civicplus_www_06142021-577_agenda_packet.pdf
│   ├── civicplus_www_06142021-577_minutes.pdf
│   ├── civicplus_www_06152021-578_agenda.html
<<< snipped >>>

This appears to stem from our handling of the meeting_id variable, which is used in Asset.download to generate the file name.

Need to either debug for this locale and/or adopt an alternate convention for standardizing file names in CivicPlusSite (and generally).

An ideal solution would be storing file artifacts based on a combination of place, agency, date of meeting, committee type, document type and document format (i.e. the file suffix). For example:

# Note, place may need more careful handling
/tmp/civic_scraper/assets/ca_belvedere/20210604_city_council_agenda_packet.pdf
/tmp/civic_scraper/assets/ca_belvedere/20210604_city_council_agenda_packet.html

It's likely that we may not have all this information available for all platforms, so we may need platform specific solutions.

Or we can go in a totally different direction and just generate unique names based on a file hash, and then use asset metadata (e.g. stored in the metadata CSV) to link given files with their unique names.