[fix] Supported hkv cpu operator base cuckoo cpu operator

[LinGeLin]

Because the underlying cuckoo gpu operator is changed to hkv operator, if a user originally uses cuckoo gpu training but does not have a gpu when serving, the inference service will fail。

Brief Description of the PR:

Fixes # (issue)

Type of change

• [x] 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:

[rhdong]

Maybe we need to open some test cases for this PR.

[LinGeLin]

Maybe we need to open some test cases for this PR.

Some cpu test cases have been added, but if need ci to run these cases, it looks like need to add another ci node, is that necessary?