Added requires_grad check for params_with_grad method

[whbldhwj]

What does this PR do?

This PR fixes the silent errors of params_with_grad when param.requires_grad=False.

Before the fix, if we have params with requires_grad=False, when we call params_with_grad method (e.g., from gradient clipping:

• the original code first checks that p.grad is not None is evaluated to False
• then it tries to evaluate p.main_grad is not None. Since p doesn't have the main_grad attribute, this line of code will error out. When the outside caller doesn't catch this exception properly, this will lead to potential bugs.

For the fix, we added requires_grad check before the rest of the code.

Before submitting

• Did you have fun?
  • Make sure you had fun coding 🙃
• Did you read the contributor guideline?
• Was this discussed/approved via a Github issue? (no need for typos, doc improvements)
  • N/A
• Did you make sure to update the docs?
  • N/A
• Did you write any new necessary tests?
  • N/A
• Did you update the changelog? (if needed)
  • N/A

PR review

Anyone in the community is free to review the PR once the tests have passed.
If we didn't discuss your PR in Github issues there's a high chance it will not be merged.