[Feature] Advanced setting to select GPU when having multiple

[exuvo]

When there are multiple good GPUs in the system it is preferable to split workloads across them.
Ex VR rendering on GPU1 and eyetracking on GPU2.

All AppendExecutionProvider_x() calls in src/Baballonia/Services/DefaultInferenceRunner.cs currently default their deviceId parameter to 0. Add an setting hidden behind the advanced menu to adjust this. Caveat this is only simple if all GPUs are from the same manufacturer.

I will try to do this myself next week and send a pull-request.

[exuvo]

Apparently all my GPU crashes are from when i run baballonia alongside vrchat so i will start trying this out this weekend.

[exuvo]

The above method worked for selecting a different GPU but my second GPU is too old for AMD to support it anymore and it does not work sufficiently with MigraphX. I might buy a cheap more recent GPU for this, we will see.

[exuvo]

As i'm no longer able to work on this and probably wont for the year (finances sucks), should i just close it?