Some new backfill package noodling for efficient network crawling

[jazware]

Just playing around for now with a more efficient package for crawling the whole network and a little tool to dump the network to a JSONL file.

The basic premise of this crawling strategy is to initialize a crawler per-PDS and walk their listRepos responses concurrently, enqueueing jobs. Then each crawler can have its own concurrency limits on a per-PDS basis for getRepo allowing you to effectively horizontally scale your network crawling without putting outsized load on any one node in the network.

Ideally you should be able to crawl the whole network in <16 hours if you have the compute and BW from it maxing out at 10 getRepos per second per PDS.