auto-chunking

on a per-dataset basis, allow user to either set chunking or auto-chunk for easier use of compression.
Chunking only takes effect when call h5f%initialize(..., comp_lvl=1) where comp_lvl may be from 1 to 9.
Otherwise, data is written contiguously for fastest write speed.
The auto-chunking algorithm was based on h5py, which has a compatible open-source license.

added call h5f%chunks() to get the chunk size of a dataset

Documented verbose and debug options

cleaned up redundant code for cleanliness and speed.

Now works on IBM Power systems as well as Intel / AMD and ARM.