Document support for HDF5 datasets with scalar and simple 1-element dataspaces for scalar specs

[rly]

HDF5 datasets can have a null (empty), scalar (singleton), or simple (multidimensional) dataspace. Datasets with simple dataspace can have shape (1,). Some systems like Igor Pro can only write simple dataspaces (https://github.com/hdmf-dev/hdmf/pull/80).

For a scalar spec, both a dataset with a scalar dataspace and one with a simple dataspace with shape (1,) are valid. See also https://github.com/hdmf-dev/hdmf/pull/80 and https://github.com/hdmf-dev/hdmf/pull/269. This should be documented in the storage documentation.

[oruebel]

Since this is specific to HDF5 it may be more appropriate to document here https://nwb-storage.readthedocs.io/en/latest/

[rly]

Yes, sorry I didn't make that clear. This repo still holds the source for the storage docs: https://github.com/NeurodataWithoutBorders/nwb-schema/tree/dev/docs/storage