Python SDK: introduce deferred garbage collection queue

[teh-cmc]

This introduces a new deferred system to clear Arrow garbage that was originally allocated in Python land.

This should fix all deadlocks/segfaults/aborts past, present and future... or not 🥲

NOTE: This lives in parallel to the already existing ALL_RECORDINGS thingy, which is still needed to avoid killing and joining threads at a bad time.

Checklist

• I have read and agree to Contributor Guide and the Code of Conduct
• I've included a screenshot or gif (if applicable)
• I have tested the web demo (if applicable):
  • Using newly built examples: app.rerun.io
  • Using examples from latest main build: app.rerun.io
  • Using full set of examples from nightly build: app.rerun.io
• The PR title and labels are set such as to maximize their usefulness for the next release's CHANGELOG

• PR Build Summary
• Docs preview
• Examples preview
• Recent benchmark results
• Wasm size tracking

[jleibs]

This feels like it might be a better implementation of the attempted workaround that existed in buffered_client:

• https://github.com/rerun-io/rerun/blob/jleibs/proper_clear_containers/crates/re_sdk_comms/src/buffered_client.rs#L185-L204

I suspect this means that code can go away and stuff can be safely dropped after packet sending instead of shuffling off to yet-another thread.

[jleibs]

Makes sense.

[teh-cmc]

This feels like it might be a better implementation of the attempted workaround that existed in buffered_client:

* https://github.com/rerun-io/rerun/blob/jleibs/proper_clear_containers/crates/re_sdk_comms/src/buffered_client.rs#L185-L204

I suspect this means that code can go away and stuff can be safely dropped after packet sending instead of shuffling off to yet-another thread.

Nice catch, it's gone 🥳