Improve provenance tracking

[weecology]

Improve provenance tracking

[ethanwhite]

More thoughts on what would be nice to have here and possible approaches:

SQLite database that stores date-time, software, and
script versions for each run of each dataset, plus a hash of an export of that
dataset to check that it hasn't been changed. Additional table (joined many to
one) that records the manipulations performed for each dataset. Raw data files
are stored with an associated identifier (probably not forever, but maybe a
few versions back). Add a view_changes function that reports the manipulations conducted and a reproduce function that uses git to checkout the
versions of the code used and then runs them on the associated raw data files.

[ethanwhite]

Possible provenance modules to use:

• https://pythonhosted.org/Sumatra/
• https://github.com/recipy/recipy

Discussion of this as part of GSoC applications:

• numfocus/gsoc#2

[henrykironde]

Added #1369