fix: Add support for fake tensors

[gs-olive]

Description

• Refactor to_numpy function to handle non-tensor inputs, avoiding fake tensor issue during compilation of constants
• Refactor multiple functions to allow for numpy data types and shapes, as a workaround for torch.Tensor instantiations being fake
• Develop new "unified" data type translation which takes any of {TRT, Numpy, Torch} data types and translates them to any other of those
• Add regression test case to elicit behavior

Fixes #1951

Type of change

• New feature (non-breaking change which adds functionality)

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]

Unified data type translator dictionary which takes a TRTDataType and translates it to a Numpy, Torch, or TRT data type of the equivalent type and precision.

[frank-wei]

Are we enable the faketensor in this test? I do not quite understand the purpose of this test file?

[gs-olive]

Based on the changes in backends.py, which remove all calls to @fake_tensor_unsupported, fake tensors will be enabled by default via Dynamo/AOT. The purpose of this test is to verify that utilities like create_constant do not instantiate Torch tensors when provided scalar inputs. For example, in the test test_lowering_mul_int below, the only op in the graph will be something like:

call_function[target=torch.ops.aten.mul.Tensor](args=(%x, 7)...)

Without the changes in this PR, the above will fail at runtime because create_constant will make a torch.Tensor for the scalar 7, and this tensor will be fake (hold no values), so when TRT goes to extract the value to make a constant tensor, the script fails.

[frank-wei]

add support for fp16?

[gs-olive]

For this particular instance, the intent was just to take Python builtins and translate them to numpy arrays. If the caller specifies an FP16 dtype in the function schema, that cast will be applied later, on line 196.