Clean up Puzzle Compress Tutorial

[LianaMikael]

What does this PR do?

This PR updates the Puzzle compress tutorial:

• add vLLM inference benchmarking
• remove verbose logs
• add more information on how to choose config parameters
• fix unused import

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 74.38%. Comparing base (1d8bd20) to head (283679e).
⚠️ Report is 2 commits behind head on feature/compress.

Additional details and impacted files

@@                 Coverage Diff                  @@
##           feature/compress     #711      +/-   ##
====================================================
- Coverage             74.47%   74.38%   -0.09%     
====================================================
  Files                   183      185       +2     
  Lines                 18416    18521     +105     
====================================================
+ Hits                  13715    13777      +62     
- Misses                 4701     4744      +43     

☔ 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.

[kevalmorabia97]

We dont want to remove it but rather guard the import so if megatron is not present, this does not raise an error. Like so:

from modelopt.torch.utils import import_plugin

with import_plugin("megatron_hooks"):
    from .megatron_hooks import *

[LianaMikael]

Fixed, thanks

[kevalmorabia97]

LGTM. Thanks for the cleanup!

[kevalmorabia97]

@LianaMikael you'll need to fix code quality errors for the markdown: https://github.com/NVIDIA/Model-Optimizer/actions/runs/20377066546/job/58558289581?pr=711#step:5:129