v0.5.0

• Augmented all HDF5 formats to make them more self-contained, so that they can be used without constantly referring back to the schema. This is done by adding most of the (previously schema-only) details into HDF5 attributes.
• Store dimnames pointers in an attribute for the sparse matrices and dense arrays . This allows writers to point to any dataset in the file that contains, e.g., feature or sample names.
• Factors in HDF5 data frames are represented as groups with codes and levels. This avoids having store a separate file for the levels.
• Tidied up the CSV data frame specification wording.
• Check that HDF5 DF column names are all non-empty strings.