refactor(rust): Don't pickle entire python function for object store keying

[nameexhaustion]

We use the memory address *const T of the Arc<T> instead.

[codecov]

Codecov Report

Attention: Patch coverage is 90.00000% with 2 lines in your changes missing coverage. Please review.

Project coverage is 80.19%. Comparing base (08732c4) to head (f83f740).
Report is 25 commits behind head on main.

Files with missing lines	Patch %	Lines
crates/polars-io/src/cloud/credential_provider.rs	80.00%	1 Missing ⚠️
crates/polars-io/src/cloud/object_store_setup.rs	93.33%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #19340      +/-   ##
==========================================
- Coverage   80.21%   80.19%   -0.03%     
==========================================
  Files        1523     1523              
  Lines      209544   209871     +327     
  Branches     2434     2434              
==========================================
+ Hits       168096   168303     +207     
- Misses      40893    41013     +120     
  Partials      555      555              

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

[ritchie46]

What do we cache here exactly? Maybe we can implement hash in combination with the pointers address. Python supports that as well.

[nameexhaustion]

What do we cache here exactly?

We are caching the object stores that we build, I believe in order to reduce DNS requests because every object store instance performs its own DNS requests - (maybe ref #11607).

[codspeed-hq]

CodSpeed Performance Report

Merging #19340 will improve performances by 28.96%

_{Comparing nameexhaustion:object-store-cred-provider-hash (f83f740) with main (eb596c9)}

Summary

⚡ 1 improvements
✅ 40 untouched benchmarks

Benchmarks breakdown

	Benchmark	main	nameexhaustion:object-store-cred-provider-hash	Change
⚡	test_groupby_h2oai_q7	2.5 ms	2 ms	+28.96%