[feat] For FP8 models, keep layers specified in `ignore_layers` in their original FP8 format

For Deepseek v32 and similar FP8 models, it is preferable to keep layers specified in `ignore_layers` (such as indexer or attn) in their original FP8 format, rather than dequantizing them to BF16.

Expected behavior:

```
AR_LOG_LEVEL=TRACE auto_round --model /models/Qwen3-8B-FP8 --ignore_layers "attn"
```
All layers within attention (matching "attn") should remain in FP8 format, not be dequantized to BF16 or float.

Depends on https://github.com/intel/auto-round/issues/1283

cc @wenhuach21 @thuang6 @xin3he