Generalizations needed in GLC layer to allow mid-year mass balance time step

_From @billsacks on April 8, 2017 13:21_

I am putting a kludge in the GLC layer (in glc_InitMod.F90) that assumes mass balance time steps always occur at the end of the year: If a non-continue run starts mid-year, I'm assuming that the mass balance time step started at the beginning of that year.

If we ever want to allow mid-year mass balance time steps, we'll have to add some code to determine the correct starting time of the current mass balance time step. Note that, in principle, this can differ from the run start time: For example, say we do five mass balance time steps per year, each 73 days. If we start a run 80 days into the year, then the start time for the current mass balance time step should be set to 73 days into the year.

_Copied from original issue: NCAR/CISM#1_