Make all Keras layers serializable

[RobinKa]

get_config() needs to return a serializable dict, right now it contains some tensors and other objects which can't be serialized

[RobinKa]

Branch: https://github.com/RobinKa/tfga/tree/feature/serializable-layers

Create layer base-class GeometricAlgebraLayer which does the following:

• Store algebra object that was passed in constructor
• get_config: only store simple python objects or numpy arrays (for blade indices etc.). Instead of storing the GeometricAlgebra object, we store the metric and later recreate the algebra.
• from_config: Recreate GeometricAlgebra from metric

All other GA layers inherit from GeometricAlgebraLayer and also have the @register_keras_serializable(package="TFGA") attribute to make them serializable without passing a custom objects dict.

There's still the issue that each layer creates their own GeometricAlgebra object from the metric. Should be solved by creating only one for each unique metric somehow.