dglc%noevolve does not restart correctly on year boundaries

[mvertens]

I have discovered that DGLC%NOEVOLVE does not restart correctly on the year boundary. This is affecting all current production runs (both CESM and NorESM) that are using DGLC%NOEVOLVE.

As a simple example - say you are doing an initial 2 year run and writing restarts every year.

• Fgrg_rofi starts of at 0
• At year 1 - Fgrg_rofi becomes 100
• Each day after year 1 - dglc continues to send 100 to the mediator
• When you get to year 2, Fgrg_rofi becomes 200 and then the model is stopped.
• What SHOULD happen when you restart again is that Fgrg_rofi should continue to be 200 for every export that is sent to the mediator. However, it is 0.

What needs to be done is to write Fgrg_rofi to dglc restart files and have those read back in upon restart. This is the same thing that happens with dice.

[jedwards4b]

Using test ERS.ne30pg3_t232.BLT1850.derecho_intel.allactive-defaultio with:

./xmlchange RUN_STARTDATE=0001-12-28
./xmlchange --append CLM_BLDNML_OPTS=-ignore_warnings

I was able to demonstrate the failed restart test. I then reran the same test with PR #306 and it passes!

[jedwards4b]

I then changed
./xmlchange GRID=a%ne30np4.pg3_l%ne30np4.pg3_oi%tx2_3v2_r%r05_g%ais8:gris4_w%null_z%null_m%tx2_3v2
to test multiple ice sheets - this also passes with #306

[jedwards4b]

It turns out that the test for multiple ice sheets did not work correctly, only one sheet is stored.

[jedwards4b]

Now fixed in #307