fix: don't use kernel layernorm on Blackwell architecture to avoid "no kernel image" error

[AdamPalaxo]

When running models with FastLayerNorm (like Llama) on Blackwell GPUs (eg. RTX 5060 Ti), the following error occurs:

CUDA Error: no kernel image is available for execution on the device /usr/src/flash-attention/csrc/layer_norm/ln_fwd_kernels.cuh 236 rank=0

This happens because the FastLayerNorm CUDA kernel is not compiled for the Blackwell compute capability (12.0 for RTX 5xxx), but the code still attempts to use it.

What this PR does

• Adds a simple compute capability check:

  major, _ = torch.cuda.get_device_capability()
  is_blackwell = major > 9

Fixes # (issue)
Fixing #3342

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you read the contributor guideline,
  Pull Request section?
• Was this discussed/approved via a Github issue or the forum? Please add a link
  to it if that's the case.
• Did you make sure to update the documentation with your changes? Here are the
  documentation guidelines, and
  here are tips on formatting docstrings.
• Did you write any new necessary tests?

Who can review?

Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.