Fix key output of networkx edge getter:

[max-zipfl-fzi]

Removed the key output, which resulted in a three-dimensional tensor, when using MultiDiGraphs.
Three-dimensional edge tuples could not be converted into the required [2,-1] COO format.

Example:
Right:
print(list(G.edges(keys=False)))
[(2, 1), (5, 4), (5, 4), (5, 4), (6, 1), (7, 1)]

Wrong:
print(list(G.edges))
[(2, 1, 0), (5, 4, 0), (5, 4, 1), (5, 4, 2), (6, 1, 0), (7, 1, 0)]

Wrong:
print(torch.LongTensor(list(G.edges)))
tensor([[2, 1, 0],
[5, 4, 0],
[5, 4, 1],
[5, 4, 2],
[6, 1, 0],
[7, 1, 0]])
Right:
print(torch.LongTensor(list(G.edges(keys=False))).t().contiguous())
tensor([[2, 5, 5, 5, 6, 7],
[1, 4, 4, 4, 1, 1]])`

[rusty1s]

Looks like CI is failing. Can you ensure that all tests in test_convert.py still work?

[max-zipfl-fzi]

I forget to check non multi-graphs:
Will add a check, whether a key should be removed or if it is not present in the first place.

[codecov-commenter]

Codecov Report

Merging #3646 (6ff72a7) into master (8630831) will decrease coverage by 0.02%.
The diff coverage is 75.00%.

@@            Coverage Diff             @@
##           master    #3646      +/-   ##
==========================================
- Coverage   81.50%   81.47%   -0.03%     
==========================================
  Files         294      294              
  Lines       14757    14734      -23     
==========================================
- Hits        12028    12005      -23     
  Misses       2729     2729              

Impacted Files	Coverage Δ
torch_geometric/utils/convert.py	85.82% <75.00%> (-0.47%)	⬇️
torch_geometric/data/temporal.py	28.57% <0.00%> (-2.68%)	⬇️
torch_geometric/graphgym/optimizer.py	72.22% <0.00%> (-1.47%)	⬇️
torch_geometric/transforms/gdc.py	78.75% <0.00%> (-0.22%)	⬇️
torch_geometric/graphgym/models/layer.py	71.59% <0.00%> (-0.17%)	⬇️
torch_geometric/data/storage.py	80.89% <0.00%> (-0.08%)	⬇️
torch_geometric/loader/graph_saint.py	97.97% <0.00%> (-0.05%)	⬇️
torch_geometric/nn/conv/rgcn_conv.py	98.51% <0.00%> (-0.03%)	⬇️
torch_geometric/nn/models/gnn_explainer.py	96.72% <0.00%> (-0.02%)	⬇️
torch_geometric/deprecation.py	100.00% <0.00%> (ø)
... and 8 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 8630831...6ff72a7. Read the comment docs.

[rusty1s]

Super. Thank you :)