Test multiple GPU-based execution providers

[brainless]

As a user, I would like to be able to understand how this software works on different GPUs now there GPU-based execution providers are available. Here are the GPUs available at the moment:

• @fbilhaut has an NVIDIA RTX 4080 on a Windows device

@brainless has:

• an NVIDIA RTX 3060 on a Linux + Windows device
• all the low cost GPUs at https://www.scaleway.com/en/gpu-instances/

Tasks

• Find out the testing method/code
• Run on Windows/Linux laptops
• Run on cloud
• Document results, perhaps on README

[fbilhaut]

Thanks @brainless for filing this issue.

Note that there is already a documentation file intended for this purpose in doc/ORT.md. I already documented the settings for CUDA under Windows there.

Some simple code to start with is available in src/examples/benchmark_gpu.rs