Time

How to get the encoding of time within Iris from the CF-netCDF style to something more palatable.

Issues with CF conventions

The following issues need clarification within the CF specification:

Does it make any sense to have time zones for non-Gregorian and/or Julian calendars.
Small amount of evidence to suggest "yes": Seth McGinnis, Jonathan Gregory
What are the "month" and "year" semantics for non-Gregorian calendars. i.e. CF-1.6 discusses the length of "year" and states it as 365.2421... but this is clearly nonsensical for non-Gregorian calendars.
Which takes precedence: the spec for UDUNITS or the implementation? i.e. What do we do about bugs in UDUNITS? For example:
Bill's "2013-02-03 9"?
"hours since 1970-01-01 00:00:00 +6" loses the time zone (could this be the fault of netCDF4-python?)
CF-1.6 states the default time zone is UTC, but does it really include leap seconds?!
What are the year-zero semantics for non-Gregorian calendars?
Small amount of evidence to suggest year-zero always exists: Jonathan Gregory
But also support for clarification: Dave Allured

Use np.datetime64/'M8' and np.timedelta64 to support the gregorian/standard and proleptic_gregorian calendars.
- NB. This will be incorrect for gregorian before 1582-10-15.
Use Time360/time360 and np.timedelta64 to support the 360-day calendar.
Add Time365, Time366 as for Time360.