Add MXFP quantization tests and configuration handling
- Implement tests for building MXFP quantization configurations, including uniform and mixed precision setups.
- Add end-to-end tests for MXFP mixed precision quantization.
- Introduce helper functions for handling MXFP source tensors and expanding block scales.
- Enhance model type resolution for deepseek_v4 and llm-compressor models.
- Ensure passthrough and dequantization behaviors are correctly validated for MXFP8 and MXFP4 sources.
Signed-off-by: Xin He <xin3.he@intel.com>
