[llvm] [mlir] [mlir] Replace MLIR_ENABLE_CUDA_CONVERSIONS with LLVM_HAS_NVPTX_TARGET (PR #93008)

[Oleksandr Alex Zinenko via llvm-commits]

https://github.com/ftynse requested changes to this pull request.

Looks even cleaner, thanks! But I suspect there is something off with how the macro is interpreted, it should be an always-defined 0/1 macro.

Note that Bazel is an optional build system for LLVM, you don't necessarily have to update it.

https://github.com/llvm/llvm-project/pull/93008