6/x mx cleanup: make NVFP4Tensor use base implements

[vkuzo]

Summary:

Refactors NVFP4Tensor to use the implements function from
AOBaseTensor.

Test Plan:

pytest test/prototype/mx_formats/ -s -x

Reviewers:

Subscribers:

Tasks:

Tags:

[vkuzo]

Stack from ghstack (oldest at bottom):

• -> 6/x mx cleanup: make NVFP4Tensor use base implements #3387
• 5/x mx cleanup: rename mx inf config to MXDynamicActivationMXWeightConfig #3386
• 4/x: mx cleanup: use kernel_preference instead of gemm_kernel_choice #3385
• 3/x mx cleanup: speed up local test suite #3383
• 2/x mx cleanup: remove pack_fp6 #3382
• 1/x mx cleanup: mxtensor: rename _block_size to block_size #3381

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/ao/3387

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

❗ 1 Active SEVs

There are 1 currently active SEVs. If your PR is affected, please view them below:

• Dr CI is temporarily not working due to API fairewall

This comment was automatically generated by Dr. CI and updates every 15 minutes.