Dropout support

[priyakasimbeg]

Make the dropout rate an argument of the model call functions such that the dropout rate can be in calls to model function. Also remove logic to read dropout rate from hyperparameter config for model initialization.
Previously, the dropout rate was set during model initialization, which led to the following issues:

1. Inconsistency between external tuning ruleset where users had the ability to set the dropout rate through initialization while self-tuning ruleset only used default dropout values.
2. No ability to change dropout value during training, which would enable "sequential hyperparameter" type submissions.

Addresses: #753

[github-actions]

MLCommons CLA bot All contributors have signed the MLCommons CLA ✍️ ✅