Fix masking ratio calculation

[guarin]

Changes

• Handle class token correctly in mask ratio calculation

Closes #1583

Fixes an issue where too many tokens were masked because the class token was included in the calculation for the number of masked tokens.

The new behavior is:

• If mask_class_token=True: Calculate number of masked tokens based on full sequence length and potentially mask class token
• If mask_class_token=False: Calculate number of masked tokens based on sequence length without the class token and never mask the class token

The original MAE code also calculates the masking ratio without the class token, see: https://github.com/facebookresearch/mae/blob/efb2a8062c206524e35e47d04501ed4f544c0ae8/models_mae.py#L130

How was it tested?

• Added unit test

[codecov[bot]]

Codecov Report

All modified and coverable lines are covered by tests :white_check_mark:

Project coverage is 81.77%. Comparing base (e0799ae) to head (843c250).

Additional details and impacted files

```diff @@ Coverage Diff @@ ## master #1585 +/- ## ======================================= Coverage 81.77% 81.77% ======================================= Files 144 144 Lines 6123 6124 +1 ======================================= + Hits 5007 5008 +1 Misses 1116 1116 ```

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.