[feature]support ragged tensor in all2all

[jq]

Description

support ragged tensor in all2all HvdAllToAllEmbedding lookup
refactor embedding variable, introduce embedding_weights interface so that dynamic embedding variable, shadow variable, hvd variable can be easily supported in the lookup and safe_embedding_lookup_sparse

Brief Description of the PR:

Fixes # (issue)

Type of change

• [ ] Bug fix
• [ ] New Tutorial
• [ ] Updated or additional documentation
• [ ] Additional Testing
• [ ] New Feature

Checklist:

• [ ] I've properly formatted my code according to the guidelines
  • [ ] By running yapf
  • [ ] By running clang-format
• [ ] This PR addresses an already submitted issue for TensorFlow Recommenders-Addons
• [ ] I have made corresponding changes to the documentation
• [ ] I have added tests that prove my fix is effective or that my feature works

How Has This Been Tested?

If you're adding a bugfix or new feature please describe the tests that you ran to verify your changes: