Documentation for io binding

Is your feature request related to a problem? Please describe.
I want to bind some inputs/outputs to gpu (output of a model to be passed to another model as input).
The documentation and examples for io binding in C# are lacking. I have found these two tests here and here.
The most information is in the C header file here but onnxruntime_c_api.h header was confusing.

It is still unclear what these functions do, and when they are needed.
For example

1. Do we need to call SynchronizeBoundInputs before session.run and then SynchronizeBoundOutputs after run?
2. Do we need to call the BindInput each time before session.run or just once and then call SynchronizeBoundInputs?
3. What happens if the bound input memory values change, do we need to call the BindInput again, or just SynchronizeBoundInputs?
4. I have an input (that changes), but has a fixed size (even though the model could handle dynamic batch size). Does that count as fixedbuffer? do I need to call BindInput each time?
5. There is an output that I would like to keep on GPU. But the rest of input and outputs I do not care to bind. Still for that one input I need to change the API call for everything, because there is no method to just bind a single output to device

System information

• ONNX Runtime version (you are using): 1.10

Describe the solution you'd like
Documentation and examples for C#