feat(graphstore): support num_nodes, enabling Tuple[FeatureStore, GraphStore] in LightningLinkData

[mananshah99]

There is still cleanup work to be done around LightningNodeData, LightningLinkData, and how we fetch num_nodes. This is simply a hotfix to have functionality working, the cleanup will happen shortly.

[codecov]

Codecov Report

Merging #5270 (b3de013) into master (df93d23) will decrease coverage by 1.75%.
The diff coverage is 96.15%.

@@            Coverage Diff             @@
##           master    #5270      +/-   ##
==========================================
- Coverage   85.08%   83.33%   -1.76%     
==========================================
  Files         337      337              
  Lines       18607    18634      +27     
==========================================
- Hits        15832    15528     -304     
- Misses       2775     3106     +331     

Impacted Files	Coverage Δ
torch_geometric/data/graph_store.py	92.13% <87.50%> (-0.55%)	⬇️
torch_geometric/data/data.py	91.35% <100.00%> (+0.08%)	⬆️
torch_geometric/data/hetero_data.py	94.54% <100.00%> (+0.08%)	⬆️
torch_geometric/data/lightning_datamodule.py	48.82% <100.00%> (+16.84%)	⬆️
torch_geometric/loader/link_neighbor_loader.py	94.70% <100.00%> (+1.62%)	⬆️
torch_geometric/testing/feature_store.py	100.00% <100.00%> (ø)
torch_geometric/testing/graph_store.py	100.00% <100.00%> (ø)
torch_geometric/nn/models/dimenet_utils.py	0.00% <0.00%> (-75.52%)	⬇️
torch_geometric/nn/models/dimenet.py	14.51% <0.00%> (-53.00%)	⬇️
torch_geometric/profile/profile.py	37.89% <0.00%> (-26.32%)	⬇️
... and 17 more

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

[mananshah99]

Completely agree that we need to stop replicating this logic across PyG. My preference is to handle this in a separate PR, and merge this one so we maintain correctness for link prediction with remote backends. The separate PR will be out soon.

[yaoyaowd]

Minor comment.