fix: add `set_fast_math` Method

[iihimanshuu]

Description:

Line 687 in llvmliteir.py calls self.set_fast_math(True) and self.set_fast_math(False), but this method is not defined in the LLVMLiteIRVisitor class. This will cause an AttributeError at runtime when vector operations with fast math are attempted.

File affected:

• llvmliteir.py

Impact: This is a critical bug that would cause runtime failures when using vector operations with fast math attributes.

[hemannt003]

Hi, I would like to work on this issue.

My plan is to implement a set_fast_math method in LLVMLiteIRVisitor to allow enabling or disabling LLVM fast-math optimizations during floating-point IR generation.

Implementation approach:
• Add a set_fast_math(self, enabled: bool) method that toggles the _fast_math_enabled flag.
• Ensure _fast_math_enabled is initialized in the constructor.
• Update floating-point operations (fadd, fsub, fmul, fdiv, and FMA) to apply fast-math flags through the existing _apply_fast_math helper.
• Make sure the flag is enabled only for the scope where fast-math is requested and reset afterward to avoid affecting unrelated instructions.

This will allow the IR builder to generate optimized floating-point instructions when the AST node requests fast-math.

If this approach looks good, I’d be happy to take this issue.