[llvm] [SDAG][NVPTX] Add TLI check for preferring custom FP_TO_SINT operations to FP_TO_UINT (PR #132470)

[Kevin McAfee via llvm-commits]

================
@@ -0,0 +1,134 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
----------------
kalxr wrote:

The first two test cases should have different PTX with the change.
If we look at line 21 in the PTX of that godbolt link:
`cvt.rzi.s16.f32         %rs1, %f1;`
Compare that to the expected output in the test, line 10:
`; CHECK-NEXT:    cvt.rzi.u16.f32 %rs1, %f1;`
The difference is the s16 vs u16. Current LLVM is generating a signed cvt instruction for what should be an unsigned value.

You can see the SASS differences here: https://godbolt.org/z/GjqYzejPz

https://github.com/llvm/llvm-project/pull/132470