Introduce IDL-based Tensor/Image APIs for python.

[jleibs]

What

Implement the assorted init / to_arrow helpers necessary to make tensors and images work.

In particular, needed to override TensorData.__init__ to provide better auto-validation and more flexibility.

Examples of valid constructions for the TensorData

TensorData(array=[[1,2,3],[4,5,6]])
TensorData(array=[[1,2,3],[4,5,6]], names=['a', 'b'])
TensorData(buffer=[1,2,3,4,5,6], shape=[TensorDimension(2, name='a'), TensorDimension(3, name='b')]

Additionally, the Tensor or Image archetypes can be derived directly from an array-type,

Tensor([[1,2,3],[4,5,6]])
Tensor(TensorData(array=[[1,2,3],[4,5,6]]))

Image([[1,2,3],[4,5,6]])
Image(TensorData(array=[[1,2,3],[4,5,6]]))

Needed to add the ability to override init for archetypes in order to patch in validation on image shapes.

Part of: #2792

Checklist

• I have read and agree to Contributor Guide and the Code of Conduct
• I've included a screenshot or gif (if applicable)
• I have tested demo.rerun.io (if applicable)

• PR Build Summary
• Docs preview
• Examples preview
• Recent benchmark results
• Wasm size tracking

[teh-cmc]

rename to image_simple.py once rebased!