[clang] [llvm] [Clang] Add env var for nvptx-arch/amdgpu-arch timeout (PR #102521)

[Joseph Huber via llvm-commits]

jhuber6 wrote:

> FWIW, on one of my test systems that's currently very busy, check-offload consistently fails with amdgpu-arch timeouts. However, the following so far makes it consistently succeed:
> 
> ```
> $ CLANG_TOOLCHAIN_PROGRAM_TIMEOUT=60 ninja check-offload
> ```
> 
> Previously, I've decreased lit parallelism to avoid the timeouts, but it hasn't worked consistently, and it slows down the test suite.

Possibly worth increasing it to a minute then, this is basically just a last ditch effort to prevent builds from hanging forever. I've had it happen personally where either the CUDA or AMDGPU drivers were stuck, and when I tried to rebuild `libc` `ninja` just hung indefinitely and it took me awhile to figure out why.

https://github.com/llvm/llvm-project/pull/102521