VertexAI compatibility layer

[VimWizardVersace]

In order to use Vertex AI and EE to host our model three things need to be done:

1. There needs to be a preprocessing layer that reshapes the input of tensors to the model to conform with how EE sends predictions in square patches. It should also output in square patches.
   There's a good example on how to do this here: (https://colab.sandbox.google.com/github/google/earthengine-community/blob/master/guides/linked/Earth_Engine_TensorFlow_Decision_Forests.ipynb#scrollTo=ncwsNhjL0pCb)
2. There needs to be a preprocessing layer that also standardizes the input.
3. There needs to be a wrapper around the models the serializes and deserializes input and output from base64 encoded tensors. There's an example of how to do this here (https://colab.sandbox.google.com/github/google/earthengine-community/blob/master/guides/linked/Earth_Engine_TensorFlow_Decision_Forests.ipynb#scrollTo=d6p_54ym0xCz)