Use HDF5 for data storage

[tmolteno]

The use of python pickle to store Observation and Visibility objects is not a good long term option. This is not robust to python2 vs python3, so persistence will move to HDF5 files.

This will add a dependency to the 'tart' python package (it will depend on h5py). Some routines will need to be written to convert existing files.

• Write new persistence methods. with tests.
• Utilities to convert from pickle to .hdf (written in python2).
• Update testbenches
• Web app save HDF RAW files
• Web app save HDF Vis files

[tmolteno]

All that is required now is to change the file extension to .hdf in tart_web_api/service.py. Testbenches are done and passing for exporting and importing lists of visibilities into .hdf5 files.