inductive_train_test_split() and split_graph() functions

[ogawayuto]

Hi,

This PR is based on the discussion in Discussion #8238. While it's possible to use masks and the subgraph() function to implement inductive node/link tasks, there's currently a lack of explicit functions for splitting data into training and testing sets specifically for inductive settings. Furthermore, when dealing with inductive link prediction, it's crucial to distinguish which edges are intended for training and testing. However, the subgraph() function doesn't provide information about test edges.

To address these limitations, I've introduced the inductive_train_test_split() function. This function streamlines the process of dividing a graph into a train graph and a test graph. It enables you to specify which edges should be allocated for training and testing, ensuring a clear separation of data for inductive tasks. It's also designed to be easily integrated with scikit-learn functions such as KFolds() or StratifiedKFolds().

This commit includes three main changes:

1. Modification of utils/__init__.py.

2. Creation of inductive_train_test_split.py.

This file contains two functions: inductive_train_test_split() and split_graph(). The former leverages the latter and operates on Data objects. It also performs checks to ensure there is no overlap between the two subsets and that all nodes are covered.

3. Creation of inductive_train_test_split_test.py.

This file includes tests for the two functions using simple examples. The test data aligns with the examples provided in the docstring.

Your reviews and comments on this contribution are greatly appreciated!

[codecov]

Codecov Report

Merging #8243 (6e89a45) into master (9f7e824) will decrease coverage by 0.60%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master    #8243      +/-   ##
==========================================
- Coverage   89.00%   88.40%   -0.60%     
==========================================
  Files         475      476       +1     
  Lines       28841    28912      +71     
==========================================
- Hits        25670    25561     -109     
- Misses       3171     3351     +180     

Files	Coverage Δ
torch_geometric/utils/__init__.py	100.00% <100.00%> (ø)
...orch_geometric/utils/inductive_train_test_split.py	100.00% <100.00%> (ø)

... and 37 files with indirect coverage changes

📣 Codecov offers a browser extension for seamless coverage viewing on GitHub. Try it in Chrome or Firefox today!

[ogawayuto]

I'm ready for the pull request review !