Support many storage formats

[ldalessa]

The tensor template currently supports two storage formats, either a tensor-managed stack-allocated row-major array format, or an externally allocated row-major array format. This is somewhat restrictive as there are many other storage formats we would like to experiment with, e.g., column-major storage, tensor-managed heap storage, external struct-of-array formats. We would also like to be able to investigate storage formats that can handle things like symmetry and anti-symmetry.