[MXFP] Fix fp4 fp4 scaled_dot does not use dpas issue

[LiyangLingIntel]

The root casue of this issue is the FP32_FP32_FP4_FP4 dot pattern is missed in DPASEngineType.
This change also adds default input precision as "tf32" when decomposing dot_scaled to dot, since dot_scaled op does not have input precision attribute.

[Copilot]

Pull Request Overview

This PR fixes an issue where FP4-FP4 scaled dot operations were not utilizing DPAS (Dot Product Accumulation Systolic) functionality. The root cause was a missing FP32_FP32_FP4_FP4 pattern in the DPAS engine type enumeration and analysis logic.

• Added the missing FP32_FP32_FP4_FP4 DPAS engine type
• Added detection logic for FP4-FP4 dot operations in DPAS analysis
• Set default input precision to TF32 when decomposing scaled dot operations

Reviewed Changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated no comments.

File	Description
third_party/intel/include/Analysis/DPAS.h	Added FP32_FP32_FP4_FP4 enum value to support FP4-FP4 DPAS operations
third_party/intel/lib/Analysis/DPAS.cpp	Added detection logic for FP4-FP4 operations and removed outdated comments
third_party/intel/lib/TritonIntelGPUTransforms/DecomposeScaledBlocked.cpp	Added default TF32 input precision when creating DotOp from DotScaledOp

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[etiotto]

Can you add a lit test ?