hotfix: top_k save policy; coat_mini & coat_tiny recipe

[geniuspatrick]

Thank you for your contribution to the MindCV repo.
Before submitting this PR, please make sure:

• You have read the Contributing Guidelines on pull requests
• Your code builds clean without any errors or warnings
• You are using approved terminology
• You have added unit tests

Motivation

If using top_k policy, the type checking of metric should be isinstance(metric, Tensor) instead of type(metric) is not np.ndarray.

In coat.yaml, drop_overflow_update: True should be added when using loss_scale_type: dynamic.

Test Plan

(How should this PR be tested? Do you require special setup to run the test or repro the fixed bug?)

Related Issues and PRs

#684
#662
#600