Sepiida

----
Modify support for timezones to be timezone aware

Any datetime going out of an API through a datetime field that does not have timezone data will be assumed to be UTC.
Previously we just ignored the data and assumed the application knew what it was doing. Now we forcibly convert it to
UTC and indicate in the ISO format that we are expressing the time in UTC.

The compliment to this is that we now take in datetime values that express a timezone and properly convert them to
UTC before handing them to the response handler functions

The major outgrowth of this is that existing code will now go from outputting datetime values such as

2017-01-11T18:31:15.133

as

2017-01-11T18:31:15.133+00:00

Which may break your unit tests

-----
Add support for sorting of sepiida endpoint fields using the 'sort' query argument

Any implementing api/platform can now use an endpoint's `sorts` property to see any requested
sorting a user made on a list request. The sort query arg supports ascending and descending sort ordering.
Ordering is ascending by default, but if a '-' is prefixed to the field name, it will be descending instead.
A single field can be provided, or multiple fields using a comma separated list.


Examples:

http://your-site.com/endpoint/?sort=foo
http://your-site.com/endpoint/?sort=foo,-bar
http://your-site.com/endpoint/?sort=-bar,foo


Note: The 2nd and 3rd example will provide different results, as ordering of the fields matters.

See http://jsonapi.org/format/fetching-sorting for more details.

-----
Fix installation on window when git is not present

-----
Update to cryptography 1.7.1 for use with windows

-----
Update permission fixture to use '*' as the holder on public permissions rather than None. Passing in None will still work
when calling grant(), however, permissions will return '*' when None is provided to the function so you may need to update
your tests accordingly

-----
Change LOGGER type for permissions and memberships. When the response to permissions and memberships fails, we simple have a LOGGER debug the error and an error being raised. This prevents from duplicate errors being raised for the same failure.