[Relax][ONNX] Fix Cast operator float->int NaN/Inf handling

[cchung100m]

Hi Committers,

This PR is trying to fix issues #19542. Any suggestions would be appreciated if you are available.

Root cause:

FP to INT lowering can be implementation-defined or UB for NaN/Inf and extreme floats, producing backend-dependent results versus ONNX Runtime.

Solution:

Apply a minimal, deterministic frontend sanitization for float to integer Casts: map NaN and ±Inf to 0.0 before astype. This prevents NaN/Inf from reaching backend fptosi/fptoui lowers and yields stable behavior across targets.

[gemini-code-assist]

Code Review

This pull request updates the ONNX frontend in TVM Relax to handle casting from floating-point types to integer types by replacing NaN and Inf values with 0.0 before performing the cast. The reviewer suggested simplifying the check for non-finite values by using relax.op.logical_not(relax.op.isfinite(src)) instead of combining isnan and isinf with a logical OR, which reduces the number of operator calls and simplifies the generated Relax graph.