Slack importer threading

[rheaparekh]

Using Zulip's run_parallel method to thread downloads.

Test:
./manage.py convert_slack_data <slack_zip_data> --token <token> --output <output_dir> --threads <threads>

[rheaparekh]

@timabbott Can you test this out?

[timabbott]

It looks like the size piece her got lost; is that intentional?

[rheaparekh]

@timabbott I removed it as I saw that it is not really required in the import process. See the function import_uploads_local.

[timabbott]

This works great (in testing, I had a 1.5-6x improvement in total import speed with this). Not sure why the results varied so much when I tested a few times. I merged this since it'll make basically every other aspect of Slack import development a lot nicer, but posted one comment above on the size attribute. I didn't see any evidence that we actually read the size field on the import size, so I don't think it's a big issue, but we should make sure we're making a sensible decision here. At the very least, I think it deserves a comment.

[rheaparekh]

@timabbott As size is not being used anywhere in the import, I removed it. I'll add a comment for it.

[rht]

As with the variation in the total import speed, could it be that the download rate after several repeated downloads is throttled by AWS?

[timabbott]

Or my local network. I don't think that's interesting; what's interesting is just that the parallelism does help.