WIP: Generate FMA loop

[etiotto]

No description provided.

[Copilot]

Pull Request Overview

This WIP (Work in Progress) PR implements loop generation for FMA (Fused Multiply-Add) operations in the TritonGPU to LLVM conversion. It introduces a new code generation mode that generates loops instead of unrolled code for dot operations.

• Adds loop-based code generation as an alternative to the existing unrolled approach
• Implements helper functions to create control flow loops using MLIR's ControlFlow dialect
• Integrates the loop generation into the existing FMA dot operation conversion pipeline

---

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

[Copilot]

Debug output statements should be removed from production code. These debugging prints will pollute the output in release builds.

Suggested change

	llvm::errs() << "at line: " << __LINE__ << "\n";
	llvm::errs() << "Module after:\n";
	mod->dumpPretty();
	llvm::errs() << "\n";