Add edge-count consistency checks in `test_graph_creation

[Shivampal157]

## Summary
The test `tests/test_graph_creation.py` verifies that all expected graph files exist and roughly checks their shapes. There is a TODO hinting that we should also verify **edge-count consistency** across files:
```python
# TODO: check that the number of edges is consistent over the files, for
# now we just check the number of features

[Shivampal157]

@sadamov, If this proposal looks good, I would be happy to open a PR for it.

[sadamov]

@Shivampal157 that would be lovely. Just a reminder that you don't have to debug the codebase wrt to the GSOC projects. But if you enjoy implementing this fix I am not stopping you. :) The TODO is still outstanding.

[Shyam-Sunder-saini]

Hi, I implemented the TODO in tests/test_graph_creation.py to verify edge-count consistency across the generated graph files.

Previously, the test only checked that the feature dimensions were correct. I extended the test to also ensure that the number of edges referenced in the adjacency matrices matches the number of feature rows in the corresponding feature tensors.

Specifically:

• For each *_index file, the number of edges is captured from edge_index.shape[1].
• For the corresponding *_features file, the test verifies that edge_features.shape[0] matches the stored edge count.
• This ensures consistency between adjacency matrices and their associated edge feature tensors.

This adds an additional integrity check for generated graphs beyond the existing shape validations.

I verified the change locally by running:

pytest tests/test_graph_creation.py

All tests pass successfully.

[Shivampal157]

@Shyam-Sunder-saini Thanks for your interest. I created this issue and I’m assigned to it, so I’ll implement the edge-count consistency checks myself and open a PR soon.

[joeloskarsson]

Hi, it looks like we now have two duplicate PRs of this fix in #301 and #396. @Shivampal157 @osten-antonio could you compare and see what is the best direction for this. The solutions seem quite different. From my quick look #301 seems to integrate a bit better with the existing test, but I only looked quickly. I think you who worked on this should have a better idea.

[Shivampal157]

@joeloskarsson , Noting there’s overlap with PR #301 for edge-count consistency. Happy to close #396 and keep #301, or vice versa ,whatever maintainers prefer

[Omnagar]

Hi, I looked through the discussion and both PRs (#301 and #396).

Both seem to be addressing the same TODO about checking edge-count
consistency between *_edge_index.pt and *_features.pt files.

From a quick look, #301 seems to integrate more directly with the
existing test structure, but I might be missing some context.

Happy to help review further or test either direction if that helps.