Add /opt/cuda to default candidates

[svenstaro]

Arch Linux uses /opt/cuda and I think it would be neat if it just worked out of the box.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[elibol]

Hi @svenstaro! Thanks for the contribution and your patience. Happy to support Arch out of the box. Do you mind having a look a the following?

1. The same candidate list is duplicated in cutile-compiler/src/cuda_tile_runtime_utils.rs (default_cuda_toolkit_candidates), where it drives runtime discovery of the tileiras binary. Without adding /opt/cuda there too, an Arch install would pass the build but still fail to locate tileiras at run time. Can you mirror the addition there (in the same position, after /usr/local/cuda)?
2. The README's "Configure Environment" section lists the default search locations — could you add /opt/cuda to that sentence as well?

[svenstaro]

Hey @elibol, sorry, didn't see your comment. I think I addressed your comments now.