Changes how vectorization is applied during fusion

[nathanielsimard]

Instead of limiting vectorization based on the most constraining factor, we read inputs multiple times when they can't match the highest vectorization factor. This way, we use the best vectorization factor for each tensor, no matter if others aren't able to.

[codecov]

Codecov Report

Attention: Patch coverage is 77.44565% with 83 lines in your changes missing coverage. Please review.

Project coverage is 82.23%. Comparing base (96839b9) to head (ecc5c80).
Report is 2 commits behind head on main.

Files with missing lines	Patch %	Lines
crates/burn-cubecl-fusion/src/on_write/io.rs	54.87%	74 Missing ⚠️
...es/burn-cubecl-fusion/src/on_write/trace/runner.rs	90.00%	6 Missing ⚠️
...ates/burn-cubecl-fusion/src/matmul/optimization.rs	66.66%	3 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2833      +/-   ##
==========================================
+ Coverage   82.21%   82.23%   +0.01%     
==========================================
  Files         853      853              
  Lines      113434   113704     +270     
==========================================
+ Hits        93265    93499     +234     
- Misses      20169    20205      +36     

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