[Optimization] Add custom NCHW to NHWC kernel for implicit GEMM

[wingertge]

Pull Request Template

Checklist

• Confirmed that run-checks all script has been executed.
• Made sure the book is up to date with changes in this PR.

Changes

Adds a custom NCHW to NHWC transpose kernel for use in implicit_gemm. This is faster than normal into_contiguous by specializing on this specific transposition.

Testing

All tests compatible with implicit_gemm pass with the new kernel, added a new test to ensure the kernel output is the same as into_contiguous.

[codecov]

Codecov Report

Attention: Patch coverage is 44.50262% with 106 lines in your changes missing coverage. Please review.

Project coverage is 82.53%. Comparing base (9c31f75) to head (8b3b0df).
Report is 6 commits behind head on main.

Files with missing lines	Patch %	Lines
...tes/burn-jit/src/kernel/conv/conv2d/layout_swap.rs	37.10%	100 Missing ⚠️
...s/burn-jit/src/kernel/conv/conv2d/implicit_gemm.rs	0.00%	3 Missing ⚠️
crates/burn-jit/src/template/base.rs	0.00%	3 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2530      +/-   ##
==========================================
- Coverage   82.59%   82.53%   -0.07%     
==========================================
  Files         827      828       +1     
  Lines      106712   106897     +185     
==========================================
+ Hits        88143    88231      +88     
- Misses      18569    18666      +97     

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

---

🚨 Try these New Features:

• Flaky Tests Detection - Detect and resolve failed and flaky tests

[AsherJingkongChen]

Why did you change entry point name?

[wingertge]

It's an update to cubecl, name is now debug_name and there's a new entrypoint_name. However, it should be "main" and not "kernel", I'm fixing that in the PR I'm currently opening.