Add similarity check to ONNX model conversions

[RunDevelopment]

Closes #2585.

This adds a new check to ensure that the ONNX models of PyTorch models produce the same outputs. This new check is options and off by default, because it requires an ONNX runtime.

I also cleaned up some other code along the way.

To the reviewer: please test this. I don't really do ONNX conversions, so I just tested it on a few models and it seemed to work.

[joeyballentine]

Errors when running in fp16 mode:

Full stack trace:

An error occurred in a Convert To ONNX node:

[ONNXRuntimeError] : 2 : INVALID_ARGUMENT : Unexpected input data type. Actual: (tensor(float)) , expected: (tensor(float16))

Input values:
• PyTorch Model: Value of type 'spandrel.__helpers.model_descriptor.ImageModelDescriptor'
• Data Type: fp16
• Opset: 17
• Verify: Yes

Stack Trace:
Traceback (most recent call last):
  File "C:\Users\joeyj\Documents\Git\chaiNNer\backend\src\process.py", line 174, in run_node
    raw_output = node.run(context, *enforced_inputs)
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\joeyj\Documents\Git\chaiNNer\backend\src\packages\chaiNNer_pytorch\pytorch\utility\convert_to_onnx.py", line 113, in convert_to_onnx_node
    verify_models(model, onnx_model_bytes, fp16)
  File "C:\Users\joeyj\Documents\Git\chaiNNer\backend\src\packages\chaiNNer_pytorch\pytorch\utility\convert_to_onnx.py", line 132, in verify_models
    onnx_out = session.run(
               ^^^^^^^^^^^^
  File "C:\Users\joeyj\AppData\Local\Programs\Python\Python311\Lib\site-packages\onnxruntime\capi\onnxruntime_inference_collection.py", line 220, in run
    return self._sess.run(output_names, input_feed, run_options)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
onnxruntime.capi.onnxruntime_pybind11_state.InvalidArgument: [ONNXRuntimeError] : 2 : INVALID_ARGUMENT : Unexpected input data type. Actual: (tensor(float)) , expected: (tensor(float16))

[RunDevelopment]

Okay, I think it should work now.

[joeyballentine]

looks like it works properly when there's a mismatch