Avoid bounds checking where it is safe to do so

[efaulhaber]

This PR and trixi-framework/TrixiParticles.jl#656 yield a ~3% speedup on a Threadripper 3990X and a ~42% speedup on an Nvidia H100. Interestingly, there is no difference on an Nvidia RTX 3090 with FP64 (probably limited by something else), but it benefits even more from inbounds on FP32 with a 60% speedup.

Not all bounds checks could be safely removed. See #79 for a completely exception-free kernel. The additional speedup is about 2% and can be neglected.

[codecov]

Codecov Report

Attention: Patch coverage is 87.50000% with 3 lines in your changes missing coverage. Please review.

Project coverage is 88.15%. Comparing base (a44f7b3) to head (4fd8741).
Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
src/neighborhood_search.jl	66.66%	2 Missing ⚠️
src/util.jl	85.71%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #74      +/-   ##
==========================================
+ Coverage   88.13%   88.15%   +0.01%     
==========================================
  Files          15       15              
  Lines         489      498       +9     
==========================================
+ Hits          431      439       +8     
- Misses         58       59       +1     

Flag	Coverage Δ
unit	88.15% <87.50%> (+0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

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