Implement a node module replacements for GNN support.

[lamyiowce]

Implementing module replacements:

• torch.Module becomes a placeholder torch fn, then a placeholder op in ONNX graph, which gets expanded to a replacement implementation.
• main replacement mechanism in daceml/onnx/nodes/replacement.py and daceml/pytorch/module_replacement.py
• implementations of GAT and GCN layers in daceml/onnx/nodes/replacement_entries.py, along with appropriate tests comparing to Pytorch Geometric implementations
• Simple benchmarking code in examples

[codecov[bot]]

Codecov Report

Merging #117 (dcfdc08) into master (ecde8b3) will decrease coverage by 51.84%. The diff coverage is 9.31%.

@@             Coverage Diff             @@
##           master     #117       +/-   ##
===========================================
- Coverage   62.69%   10.85%   -51.84%     
===========================================
  Files          65       68        +3     
  Lines        7239     7561      +322     
===========================================
- Hits         4538      820     -3718     
- Misses       2701     6741     +4040     

Impacted Files	Coverage Δ
daceml/onnx/converters.py	24.77% <0.00%> (-71.42%)	:arrow_down:
daceml/onnx/nodes/__init__.py	0.00% <0.00%> (-100.00%)	:arrow_down:
daceml/onnx/nodes/replacement_entries.py	0.00% <0.00%> (ø)
daceml/onnx/op_implementations/__init__.py	0.00% <0.00%> (-100.00%)	:arrow_down:
...ml/onnx/op_implementations/pure_implementations.py	47.53% <ø> (-26.43%)	:arrow_down:
.../op_implementations/replacement_implementations.py	0.00% <0.00%> (ø)
daceml/torch/module.py	0.00% <0.00%> (-89.96%)	:arrow_down:
daceml/onnx/nodes/replacement.py	0.82% <0.82%> (ø)
daceml/onnx/onnx_importer.py	54.05% <14.71%> (-40.72%)	:arrow_down:
daceml/onnx/shape_inference/shape_inference.py	62.22% <59.52%> (-37.78%)	:arrow_down:

... and 58 files with indirect coverage changes

:mega: We’re building smart automated test selection to slash your CI/CD build times. Learn more