WIP: client: Set pool_max_idle_per_host to 0 on hyper::Client

[andybalaam]

NOT READY FOR REVIEW. Marked as such to encourage GH to run the relevant job.

This is just a test to see whether the errors we were seeing here: https://github.com/matrix-org/matrix-rust-sdk/actions/runs/8388525873/job/22972906570?pr=3252#step:9:1818 are fixed by setting pool_max_idle_per_host to 0 on the hyper client.

It's a total guess, but here are the issues I found that suggest it might help:

• work around hang issue in hyper Azure/azure-sdk-for-rust#1550
• Unexplained race condition in v0.16 causing "runtime dropped the dispatch task" fussybeaver/bollard#387

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 83.59%. Comparing base (97959bb) to head (d7466e3).
Report is 3 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #3269      +/-   ##
==========================================
- Coverage   83.63%   83.59%   -0.05%     
==========================================
  Files         239      238       -1     
  Lines       24686    24575     -111     
==========================================
- Hits        20647    20543     -104     
+ Misses       4039     4032       -7     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[bnjbvr]

Thanks for digging into it. I think the runtime dropped was a sign that another task timed out; so it's really the reporting that was misleading here, and as long as we all collectively remember that a timeout elsewhere in the test may cause this "dropped runtime blah blah" errors, we should be fine.

In the meanwhile, I suppose we can avoid this, and revive it if we figure out with utter confidence that it will fix the testing issue we were observing.

[andybalaam]

Great, thank you for figuring this out.