Add TorchScript support for RECT_L

[ftxj]

This PR add the TorchScript support for RECT_L model.

The fail reason and our solution for original code is very similar with PR #6721, except that this model using the torch.jit.export on embed and get_semantic_labels methods. And another fail reason is @torch.no_grad, which bring some error msg I cann't understand.

Adding TorchScript support will bring a lot of extra code and reduce the code readability. I will consider how to do better in another PR.

[codecov]

Codecov Report

Merging #6727 (e230ce8) into master (c45f8db) will increase coverage by 0.00%.
The diff coverage is 100.00%.

❗ Current head e230ce8 differs from pull request most recent head 6358fe7. Consider uploading reports for the commit 6358fe7 to get more accurate results

@@           Coverage Diff           @@
##           master    #6727   +/-   ##
=======================================
  Coverage   89.20%   89.20%           
=======================================
  Files         425      425           
  Lines       23076    23084    +8     
=======================================
+ Hits        20585    20593    +8     
  Misses       2491     2491           

Impacted Files	Coverage Δ
torch_geometric/nn/models/rect.py	100.00% <100.00%> (ø)

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

[rusty1s]

Thanks. I like it. Your comment about code duplication makes a lot of sense to me, although I am not entirely sure how to do better here. Not sure if we can elegantly auto-generate these wrapper classes.