Handle process names in the json creation script about areweslimyet

[Nephyrin]

create_graph_json.py handles existing metadata by adding it to the datapoint, e.g.

Meta	Datapoint
Iteration 1:CheckpointFoo	blah/explicit/js

becomes Iteration 1/CheckpointFoo/blah/explicit/js

The reason these fields aren't just baked into the datapoint by the test to begin with is the silly string de-duplication stuff. We don't want to store each reporter name iterations * checkpoints times per run in sqlite. A database with proper transparent compression would solve this, so we could just prefix whatever onto the reporter name and it would work.

Also, the website is hardcoded to mark certain 'depths' of the tree as special in a few spots, but these are easy to increment if we add another level for process:
https://github.com/mozilla/areweslimyet/blob/master/html/slimyet.js#L820
https://github.com/mozilla/areweslimyet/blob/master/html/slimyet.js#L821
https://github.com/mozilla/areweslimyet/blob/master/html/slimyet.js#L928

[Nephyrin]

For the record, if you change the JSON format, the procedure to re-export the old databases is roughly:

# Run this over a weekend, it wont be quick, these DBs are huge and sqlite slow
for compressedMonthDB in db/areweslimyet-*.sqlite.xz; do
  monthDB="${compressedMonthDB%.xz}"
  seriesName="$(basename "${monthDB%.sqlite}")"
  monthJSON="html/data/$seriesName.json.gz"
  util/unarchive_db.sh "$compressedMonthDB"
  util/update_db.py "$monthDB"
  mv -v "$monthJSON" "$monthJSON.oldFormat"
  ./create_graph_json.py "$monthDB" "$seriesName" html/data/
  util/archive_db.sh "$monthDB"
done

(do this one at a time with a script because having them all uncompressed at once will fill the disk)

[EricRahm]

As of ae88a34 create_graph_json.py handles the new DB format but doesn't actually use anything but the main process. The next step is to update the script to use the process names (and annotate kinds).

[EricRahm]

This landed in a11f3ec on the e10s branch.