- nyr, month, mday, sec are the new time manager prognostic variables
- time is renamed to timesec and is a diagnostic, not prognostic variable
- remove time_forc, was not needed or being used anyway
- refactor ice_calendar, use calendar_type to define calendar after initialization
- add several new subroutines to ice_calendar including
  - advance_time = handles advancing the model a timestep
  - calendar_compute_elapsed_days = compute elapsed days since 0000-01-01
  - calendar_compute_days_between = compute days between two yyyy-mm-dd dates
  - calendar_set_date_from_timesecs = support conversion of timesecs to date for binary restarts
  - calendar = same behavior as old method, just no more argument
- add month_init, day_init, sec_init, and npt_unit to namelist
- this is bit-for-bit with f5f487f on cheyenne with full test suite, manual checks of log files
  were done to verify.  restart files are not identical due to changes in files saved
- old restarts should be backwards compatible
- to advance the model a timestep, call advance_time() or update nyr, month, mday, sec and call calendar().
  subroutine calendar() will adjust the date values as needed (for instance, adding a million seconds to sec).

Modify set_nml.run* to use new npt_unit and npt
Add calc_timesteps test to calchk unit test
Update standalone model to call calc_timesteps during initialization
  and after restarts are read

Update computation of fyear to support cycling correctly with new definition on nyr
Modify restart run lengths to be days instead of timesteps
Add debugging to ice_forcing.F90
Add timer for forcing

- Update drivers/mct/cesm1/CICE_InitMod.F90 and CICE_RunMod.F90
  - use advance_timestep
  - get rid of use of time variable, shift to nyr, month, mday, sec
  - update calendar call
  - update usage of year_init
- Add section in user guide about coupling with new time manager
- Fix stuff unrelated to time manager
  - Fix bug in wght_file diagnostic in ice_domain.F90, was written before defined
  - Fix bug in bathymetry settings when max kmt is greater than hardcoded 40 levels
  - Add a few missing namelist variables recently added to documentation

They were nyr, month, mday, sec.  The "m" in front means model and it
creates more unique names in the source code than year, month, day, sec
which can be hard to grep for because there are so many matches.

This is bit-for-bit on a full test suite on cheyenne.