[Metal] Abort in patchIsSpecialFloat for vectorized isnan or isinf

Nobody has claimed this yet.

We are seeing the failure on two tests:
> OffloadTest-mtl :: Feature/HLSLLib/long-vector/isinf.long-vector.32.test
> OffloadTest-mtl :: Feature/HLSLLib/long-vector/isnan.long-vector.32.test

This seems to be specific to DXC and not clang-dxc which makes me think `patchIsSpecialFloat` can't handle longer vectors.
For clang-dxc it is likely not erroring because we scalarize.

https://feedbackassistant.apple.com/feedback/24450254

It looks like patchIsSpecialFloat might not have the ability to handle vector floats. I am not on the latest version of metal-shaderconverter.

metal-shaderconverter --version
metal-irconverter version: 3.0.0

```gdb
# | metal-shaderconverter: [patchIsSpecialFloat:7420] Fatal: patchIsSpecialFloat argument was not a valid floating point type
# | Stack dump without symbol names (ensure you have llvm-symbolizer in your PATH or set the environment var `LLVM_SYMBOLIZER_PATH` to point to it):
# | 0 offloader 0x0000000104a39208 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) + 56
# | 1 offloader 0x0000000104a36f50 llvm::sys::RunSignalHandlers() + 204
# | 2 offloader 0x0000000104a39d08 SignalHandler(int, __siginfo*, void*) + 320
# | 3 libsystem_platform.dylib 0x000000018a843744 _sigtramp + 56
# | 4 libsystem_pthread.dylib 0x000000018a839888 pthread_kill + 296
# | 5 libsystem_c.dylib 0x000000018a73e850 abort + 124
# | 6 libmetalirconverter.dylib 0x0000000108ec0ff0 AIRBuilder::patchIsSpecialFloat(llvm::MDNode*, llvm::Instruction&, std::__1::basic_string, std::__1::allocator>&) + 1172
# | 7 offloader 0x0000000104a73c7c (anonymous namespace)::MTLDevice::createPipelineCs(llvm::StringRef, offloadtest::BindingsDesc const&, offloadtest::ShaderContainer) + 212
# | 8 offloader 0x0000000104a732c4 (anonymous namespace)::MTLDevice::executeProgram(offloadtest::Pipeline&) + 3868
# | 9 offloader 0x00000001049dc998 main + 2072
# | 10 dyld 0x000000018a471d54 start + 7184
# `-----------------------------
# error: command failed with exit status: -6
```

## Example cases:

```hlsl
StructuredBuffer In0 : register(t0);
RWStructuredBuffer Out : register(u1);

[numthreads(1,1,1)]
void main() {
vector Input0_5;
for (uint I = 0; I < 5; ++I)
Input0_5[I] = In0[0 + I];
vector Result5 = isnan(Input0_5);
for (uint I = 0; I < 5; ++I)
Out[0 + I] = Result5[I];
}
```

```hlsl
StructuredBuffer In0 : register(t0);
RWStructuredBuffer Out : register(u1);

[numthreads(1,1,1)]
void main() {
vector Input0_5;
for (uint I = 0; I < 5; ++I)
Input0_5[I] = In0[0 + I];
vector Result5 = isinf(Input0_5);
for (uint I = 0; I < 5; ++I)
Out[0 + I] = Result5[I];
}
```

Start with the failing OffloadTest-mtl long-vector/isinf.long-vector.32.test and isnan.long-vector.32.test cases, then inspect the patchIsSpecialFloat entry point named in the stack trace. Reproduce the abort with the provided HLSL examples and compare DXC with clang-dxc. Done means both vectorized tests complete without the invalid floating-point type abort.

Written by the indexing model from the issue text.