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:

flux.1-dev with 2048x2048 image size

Config	Median Time (s)	Speedup
bf16 baseline + torch.compile	3.17	1.00x
fp8_attn + compile	2.66	1.07x

Wan2.1-14B-Diffusers with 1280x720 frame size and 81 frames

Config	Median Time (s)	Speedup
bf16 baseline + torch.compile	81.32	1.00x
fp8_attn + compile	46.88	1.18x

Note that these results use a naive scheme, where every layer is quantized. Doing so results in subpar results when doing many inference steps (50 for WAN2.1). Using a better scheme (such as skipping early and late layers, quantizing 36/40 total layers) results in 1.12x speedup with much better quality. The VBench benchmarks for that are below: