fix: Repair output binding indexing scheme in TRT

[gs-olive]

Description

• Output binding indices are not guaranteed to be strictly after input binding indices in value
• Instead, we can use the output binding map as the ground truth for the mapping from TRT indices to PyTorch Tensor indices
• This bug is encountered when using TRTModuleNext on certain models (see issue below)

Fixes #2053
Addresses Second Bug in #1565

Type of change

• Bug fix (non-breaking change which fixes an issue)

Checklist:

• [ x ] My code follows the style guidelines of this project (You can use the linters)
• [ x ] I have performed a self-review of my own code
• [ x ] I have commented my code, particularly in hard-to-understand areas and hacks
• [ x ] I have made corresponding changes to the documentation
• [ x ] I have added tests to verify my fix or my feature
• [ x ] New and existing unit tests pass locally with my changes
• [ x ] I have added the relevant labels to my PR in so that relevant reviewers are notified

[gs-olive]

In the case of the detectron model in #2053, the output binding map was of the form:

DEBUG: [Torch-TensorRT - Debug Build] - Input binding name: arg264_1 has TensorRT binding index: 0, Torch binding index: 0
DEBUG: [Torch-TensorRT - Debug Build] - Input binding name: arg268_1 has TensorRT binding index: 1, Torch binding index: 1
DEBUG: [Torch-TensorRT - Debug Build] - Input binding name: add_121 has TensorRT binding index: 2, Torch binding index: 2
DEBUG: [Torch-TensorRT - Debug Build] - Input binding name: clamp_51 has TensorRT binding index: 3, Torch binding index: 3
DEBUG: [Torch-TensorRT - Debug Build] - Input binding name: arg72_1 has TensorRT binding index: 7, Torch binding index: 4
DEBUG: [Torch-TensorRT - Debug Build] - Input binding name: arg73_1 has TensorRT binding index: 8, Torch binding index: 5
DEBUG: [Torch-TensorRT - Debug Build] - Input binding name: arg74_1 has TensorRT binding index: 9, Torch binding index: 6
DEBUG: [Torch-TensorRT - Debug Build] - Input binding name: arg75_1 has TensorRT binding index: 10, Torch binding index: 7
DEBUG: [Torch-TensorRT - Debug Build] - Output binding name: output0 has TensorRT binding index: 4, Torch binding index: 8
DEBUG: [Torch-TensorRT - Debug Build] - Output binding name: output1 has TensorRT binding index: 5, Torch binding index: 9
DEBUG: [Torch-TensorRT - Debug Build] - Output binding name: output2 has TensorRT binding index: 6, Torch binding index: 10
DEBUG: [Torch-TensorRT - Debug Build] - Output binding name: output3 has TensorRT binding index: 11, Torch binding index: 11

Some of the output bindings have TRT indices which are smaller than the number of input bindings, causing errors when the out_binding_map is accessed with a fixed index like o, as above.

[narendasan]

LGTM