Add PyTorch SparseTensor support for GINConv, SAGEConv, and GraphConv

[EdisonLeeeee]

This PR aims to add PyTorch SparseTensor support for GINConv, SAGEConv, and GraphConv, and also update type hints in message_and_aggregate correspondingly.

[EdisonLeeeee]

Testing failed because I changed the type hints from

def message_and_aggregate(self, adj_t: SparseTensor, x: Tensor) -> Tensor:

to

def message_and_aggregate(self, adj_t: Adj, x: Tensor) -> Tensor:

@rusty1s Any solution?

[rusty1s]

I think we should keep them for now since we don't support JIT for torch.sparse.Tensor anyway yet.

[codecov]

Codecov Report

Merging #6532 (216fa11) into master (6d6162d) will increase coverage by 0.00%.
The diff coverage is 100.00%.

@@           Coverage Diff           @@
##           master    #6532   +/-   ##
=======================================
  Coverage   85.75%   85.75%           
=======================================
  Files         406      406           
  Lines       22725    22730    +5     
=======================================
+ Hits        19487    19492    +5     
  Misses       3238     3238           

Impacted Files	Coverage Δ
torch_geometric/nn/conv/gin_conv.py	97.40% <100.00%> (+0.06%)	⬆️
torch_geometric/nn/conv/graph_conv.py	100.00% <100.00%> (ø)
torch_geometric/nn/conv/sage_conv.py	96.87% <100.00%> (+0.10%)	⬆️

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