Fix gradient norm tracking and gradient clipping

[carmocca]

What does this PR do?

Fixes #9599, #9330, #9205, #9694

Before:

closure:
    zero_grad()
    forward()
    backward()
    <-- MISSING UNSCALE
    track_norm()
    clip_grad()
optimizer.step()
    closure()
    unscale()
    actual_step()

After:

closure:
    zero_grad()
    forward()
    backward()
optimizer.step()
    closure()
    unscale()
    actual_step()
track_norm()
clip_grad()

Zero will happen on the next step so the gradients will not have been zero-ed out

Does your PR introduce any breaking changes? If yes, please list them.

None

Before submitting

• Was this discussed/approved via a GitHub issue? (not for typos and docs)
• Did you read the contributor guideline, Pull Request section?
• Did you make sure your PR does only one thing, instead of bundling different changes together?
• [n/a] Did you make sure to update the documentation with your changes? (if necessary)
• Did you write any new necessary tests? (not for typos and docs)
• Did you verify new and existing tests pass locally with your changes?
• Did you list all the breaking changes introduced by this pull request?
• Did you update the CHANGELOG? (not for typos, docs, test updates, or internal minor changes/refactorings)

PR review

• Is this pull request ready for review? (if not, please submit in draft mode)
• Check that all items from Before submitting are resolved
• Make sure the title is self-explanatory and the description concisely explains the PR
• Add labels and milestones (and optionally projects) to the PR so it can be classified

[tchaton]

LGMT !