Fix crawl workflow statistics

[tw4l]

Fixes #3011

• Fix logic that previously incremented workflow's crawlSuccessfulCount for every crawl, regardless of successful or failed
• Ensure that crawl files are only added to a workflow's size when crawl successfully completes
• Fix handling of crawl deletion to account for whether crawl was successful or not
• Add migration to recalculate stats for all workflows
• Fix stats_recompute_all so that it actually works as expected - it previously failed because we were attempting to call get_running_crawl on the crawl_configs mongo collection rather than on CrawlConfigOps

Testing

1. Spin up a local instance of Browsertrix on main
2. Create a workflow and run several crawls. Have some of these crawls be successful, have others be failed or canceled, including canceled after pausing where WACZ files have already been uploaded
3. Go to Network tab and verify the statistics for the workflow are incorrect, including that all crawls are counted as successful and the size includes the WACZs from paused crawls that then failed/were canceled
4. Switch to this branch and run the new migration
5. Go back to the Network tab and verify the statistics for the workflow now match what we expect

[ikreymer]

Looks good! Tested locally added additional clearing for when there are no crawls, as found a case (without this branch) where crawlSucessfulCount was -1, but there were no crawls, so the stats were not being cleared, and now are!