Add XENetConv convolution layer

[minirawat]

As requested in issue tracker #8257 (#8257),
this PR implements the XENet Convolution layer as outlined in the paper:

  > [XENet: Using a new graph convolution to accelerate the timeline for protein design on quantum computers](https://www.biorxiv.org/content/10.1101/2021.05.05.442729v1)
  > Jack B. Maguire, Daniele Grattarola, Eugene Klyshko, Vikram Khipple Mulligan, Hans Melo

This PR implements:

• A XENetConv convolution layer as outlined in the above paper for PyG framework
• Set of test cases (basic) to validate the implementation
• I have tested this with Cora dataset and was able to achieve an Test accuracy of 53% with my
  base hyperparameters. I could see the accuracy go up to 63% by manually experimenting
  with different hyperparameter values.
  • I did try testing this with the ZINC dataset as this is probably a more realistic dataset for this model
    as discussed here (Graph classification with node and edge features benchmark #2198). However, I am
    running into issues with the ZINC data processing step in PyG. I’ll investigate this as a follow up and
    will file any relevant issue tickets to PyG, as needed.
• I will also follow up with an example implementation with the Cora dataset as a separate PR.
• Ran pytest and all test cases pass
• All linting checks pass

This was done as part of the course work for CS224W.

Benchmark from the Cora run:

Epoch 010, Loss: 1.7223, Val Acc: 0.2220, Test Acc: 0.2140
Epoch 020, Loss: 1.6110, Val Acc: 0.2520, Test Acc: 0.2670
Epoch 030, Loss: 0.6450, Val Acc: 0.4060, Test Acc: 0.3900
Epoch 040, Loss: 2.5172, Val Acc: 0.2460, Test Acc: 0.2500
Epoch 050, Loss: 0.5935, Val Acc: 0.3260, Test Acc: 0.3250
Epoch 060, Loss: 0.0863, Val Acc: 0.4620, Test Acc: 0.4710
Epoch 070, Loss: 0.0225, Val Acc: 0.4700, Test Acc: 0.4590
Epoch 080, Loss: 0.0136, Val Acc: 0.4520, Test Acc: 0.4700
Epoch 090, Loss: 0.0048, Val Acc: 0.4780, Test Acc: 0.4970
Epoch 100, Loss: 0.0022, Val Acc: 0.4740, Test Acc: 0.4830
Epoch 110, Loss: 0.0012, Val Acc: 0.4700, Test Acc: 0.4780
Epoch 120, Loss: 0.0009, Val Acc: 0.4600, Test Acc: 0.4800
Early stopping at epoch 121
Best Test Accuracy: 0.5260