[FEA] Allow io::WriteMAT to write multiple tensors.

[nvjonwong]

Currently io::WriteMAT can only write one tensor per file.

Proposal: Allow users to add to a pybind11::dict and then add tensors to that. Provide WriteMAT overload that takes a pybind11::dict directly and writes it to file.

https://github.com/NVIDIA/MatX/tree/nvjonwong/write_multiple_to_mat