[Bug #193] fix fp8 blockwise real quantization

[meenchen]

What does this PR do?

Type of change: bug fix

Overview: instead of using amax, we should compute block_scale on the fly

Usage

# Add a code snippet demonstrating how to use this

Testing

Before your PR is "Ready for review"

• Make sure you read and follow Contributor guidelines and your commits are signed.
• Is this change backward compatible?: Yes/No
• Did you write any new necessary tests?: Yes/No
• Did you add or update any necessary documentation?: Yes/No
• Did you update Changelog?: Yes/No

Additional Information

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 74.64%. Comparing base (32d168c) to head (86546cd).
⚠️ Report is 121 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #522      +/-   ##
==========================================
+ Coverage   73.38%   74.64%   +1.26%     
==========================================
  Files         180      183       +3     
  Lines       17934    18547     +613     
==========================================
+ Hits        13160    13844     +684     
+ Misses       4774     4703      -71     

☔ 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.