[Feature Request] Nvidia TensorRT RTX runtime in C#

### Describe the feature request

The option to enable the TensorRT RTX runtime in C# does not exist.
DirectML is too slow, the other options aren't portable

### Describe scenario use case

I want to ship my inference application to my customers, have them run the network at max speed, but not ship the 6GB DLL's of CUDA and CUDNN.

The issue names no files, tests, or entry points. Start by locating the C# TensorRT execution-provider integration and how runtime selection currently works; done means the TensorRT RTX runtime can be enabled from C# without shipping CUDA or CUDNN DLLs.

Written by the indexing model from the issue text.