[metal] Dynamic Threadgroup memory

[christiangnrd]

Goes with JuliaGPU/Metal.jl#750.

This approach is only supported in macOS 15+, but it felt like the best approach for me to start with to learn about how this stuff all works.

Not yet functional, I'm not sure how to get the type with the opaque pointers (or if that's even how I would need to do it)

[codecov]

Codecov Report

❌ Patch coverage is 30.30303% with 23 lines in your changes missing coverage. Please review.
✅ Project coverage is 76.87%. Comparing base (8c79865) to head (bb50223).

Files with missing lines	Patch %	Lines
src/metal.jl	30.30%	23 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #768      +/-   ##
==========================================
+ Coverage   76.75%   76.87%   +0.11%     
==========================================
  Files          25       25              
  Lines        4405     4410       +5     
==========================================
+ Hits         3381     3390       +9     
+ Misses       1024     1020       -4     

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

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.