0.3.3

• Fixed tests on Windows.
• Fixed reading and writing string arrays of length non-divisible by a power of 2.

0.3.2

• Fixed reading large (> 2G) NPY and NPZ files.
• Ensured the produced NPY file has no trailing gibberish.
• Changed NpyFile.write to automatically choose the appropriate
  format version.

0.3.1

• Changed NpyFile.write and NpzFile.Writer.write to work in
  constant amount of memory.
• Allowed writing arrays using a non-native byte order.
• Dropped Guava dependency.

0.3.0

• Added support for reading and writing arbitrary-dimensional arrays.

0.2.1

• Fixed metadata in NPY files created by NpyFile.write. See issue #4.

0.2.0

• Unified reading and writing API of NpzFile.
• Added NpzFile.Reader.introspect for exploring the metadata of an
  NPZ file without reading whole arrays.
• Fixed NpyFile behaviour for arrays with strings of different lengths.

0.1.0

• Sketched reading and writing API. This is not yet finalized and can
  change in the future releases.
• Only 1D non-structured arrays of types 'b', 'i', 'u', 'f' and 'S'
  are supported.