Support optimizers from tf.keras.optimizers in Tensorflow 2.11+

[danielenricocahall]

Describe the bug

Currently, if someone supplies an optimizer from tf.keras.optimizers.Optimizer in Tensorflow 2.11+ to SparkModel, we get the following error

self = <pyspark.cloudpickle.cloudpickle_fast.CloudPickler object at 0x7fe7fc26a1c0>
obj = (<function RDD.mapPartitions.<locals>.func at 0x7fe80bb3a790>, None, BatchedSerializer(CloudPickleSerializer(), 10), AutoBatchedSerializer(CloudPickleSerializer()))

    def dump(self, obj):
        try:
>           return Pickler.dump(self, obj)
E           TypeError: cannot pickle 'weakref' object

This can be resolved by changing the optimizer import to tf.keras.optimizers.legacy. However, we should try to support both optimizers long term.
To Reproduce
Go to tests/integration/test_end_to_end.py, and change from tensorflow.keras.optimizers.legacy import SGD to from tensorflow.keras.optimizers import SGD

Expected behavior
We should not receive an error when supplying an optimizer imported from the optimizers package.
Desktop (please complete the following information):

• OS: [e.g. iOS]
• Version [e.g. 22]

Additional context
Relevant PR: #25