Default to TensorKind.SPARSE_COO for sparse tensors

[gsakkis]

Currently we default to TensorKind.SPARSE_CSR for 2D sparse arrays on Pytorch, However the CSR sparse tensor format:

• is not implemented for Tensorflow
• is in beta state for Pytorch even on the latest version (1.12) and triggers UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues

This small PR changes the default sparse tensor kind to be always TensorKind.SPARSE_COO for both Pytorch and Tensorflow, regardless of the array shape. The user can still select CSR tensors by passing explicitly tensor_kind=TensorKind.SPARSE_CSR to ArrayParams.

[gsakkis]

Added a second commit that was enabled by the first: simplify ArrayParams.to_tensor_schema() by removing the framework-specific tensor transformation logic.