Support combination quant and cache in diffusion model

[yghstill]

• Combination strategy can setting in yaml such as:

compression:
  name: [Cache, PTQ]
  cache:
    name: DeepCache
    ...
  quantization:
    name: fp8_static
    bits: 8
    ...

• And run combin inference can execute the following command:

python tools/infer.py --model-path ./output/flux-1-schnell_deepcache_fp8_static \
        --input-prompt "A beautiful landscape with mountains and a river."