[Feat] Add torch.compile support

[felixdittrich92]

This PR:

• Add torch.compile basic support (different backends / fullgraph is up to the user - we support only the basic compile compatibility with inductor (default) backend)
• Add corresponding tests and a documentation section
• No benchmarks added here because this scales with the users hardware

Any feedback is welcome 🤗

Closes: #1684 #1690

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 96.59%. Comparing base (72bed41) to head (b16b9ad).

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1791      +/-   ##
==========================================
+ Coverage   96.57%   96.59%   +0.01%     
==========================================
  Files         165      165              
  Lines        7892     7929      +37     
==========================================
+ Hits         7622     7659      +37     
  Misses        270      270              

Flag	Coverage Δ
unittests	96.59% <100.00%> (+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.

---

🚨 Try these New Features:

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

[felixdittrich92]

@odulcy-mindee This slows down our pytorch test CI extremly (before ~9min | now ~20min) because the compilation takes some time :/

[felixdittrich92]

post proc can't be compiled so model runs fully compiled and all the other parts keept as is

[felixdittrich92]

Then i think rdy to review :)

I would say we shouldn't add a benchmark to the docs here because this is really user hardware depending or how do you see it ?

At the end the Onnx way is still to prefer ... compile brings ~5-10% boost ..Onnx boosts by +50% (on CPU) ^^

[felixdittrich92]

@odulcy-mindee Anything to add ? Otherwise ready to review ^^