Consider updating the file formats to JSON

[ocrickard]

Most programs nowadays have settled on JSON as an intermediate form for most data. Data for SHAKE is not enormous, and would be much easier to parse/prepare for other programs if it were in JSON. This comes at the price of being a little harder to read by a human, but not that much worse.

There are now several good JSON implementations in FORTRAN. We should use one. A starting

The hard part will probably be all of the assumptions in the program about data format in the subroutines. I'm honestly not sure where problems will crop up.