Add low precision attention API from torchao to TorchAoConfig

[howardzhang-cv]

What does this PR do?

Adds low precision attention API from TorchAO to diffusers by updating TorchAoConfig with attn_backend option.
Note: this will require torchao 0.17.0

Todo:

• Need to add new necessary tests
• Need to update the documentation:
  • documentation guidelines
  • here are tips on formatting docstrings.

Results:

Results are on flux.1-dev with 2048x2048 image size

Config	Median Time (s)	Speedup
bf16 baseline	4.39	1.00x
fp8_attn	4.14	1.06x
bf16 baseline + torch.compile	3.17	1.00x
fp8_attn + compile	2.66	1.19x