First version in the preparation for pull request. There is still man…

…y small improvements to be added. This version will be used as base.

Expand crop reproductive pools; remove some unused options

The main change in this tag is adding placeholder pools (and
corresponding fluxes) that will be needed to couple the AgSys crop
model: Rather than a single grain pool, AgSys has two different grain
pools (meal & oil) as well as two different reproductive structure pools
(pod & floret). This tag adds a dynamically-allocated second dimension to
the relevant variables to accommodate these new pools. The new
CropReprPoolsMod controls the allocation and naming of these pools, and
facilitates looping over the new dimension. Currently parameters in
there are hard-coded for the current crop model, but this will become
more dynamic with the coupling of AgSys.

This tag also removes experimental / untested options in
NutrientCompetitionFlexibleCNMod. Although this is mostly unrelated to
the main change, doing this simplified the changes needed for the main
change.

This tag also resolves ESCOMP#1666 (Isotopic values of grain_mr not
set correctly when use_grainproduct is false).

Finally, this tag removes or modifies some of the most expensive tests
in the aux_clm test suite to reduce the cost of running the test suite.

FATES parameter file updated to align with clm pft optical parameters

This tag incorporates updates to the FATES parameter file which now align the FATES
PFT optical parameters with the CLM PFT optical parameters.  These parameters are
sourced from the recommended parameters in Majasalmi & Bright 2019.

Update to ctsm5.1.dev054

Some changes to matrix because soil decomp CN option removed.

Implement PCT_URBAN_MAX to minimize dynamic urban memory

Read in 'PCT_URBAN_MAX' from the landuse timeseries file (maximum urban
percentage throughout timeseries) and initialize urban landunits in
memory only where PCT_URBAN_MAX is greater than zero.

 Allow user to replace vegetation in fsurdat files with any pft/cft u…

…sing the

 fsurdat_modifier tool option dom_plant. This option replaces now obsolete
 option dom_nat_pft which handled pfts only and not cfts.

Do not subtract irrigation from QRUNOFF diagnostic

Remove code that subtracts surface irrigation flux from qflx_runoff_col.
This is a diagnostic change only.

This subtraction no longer makes sense now that irrigation is passed as
a separate flux to the ROF model.

Use avg days per year when converting param units

When converting parameter units from per-year to per-second, use average
days per year instead of current number of days per year. This is
relevant when running with a Gregorian calendar.

See ESCOMP#1612 for details.

Changes to CropPhenology timing

Changes to CropPhenology timing to support
ESCOMP#1616 and other work being done by
Sam Rabin:

(1) Change CropPhenology to look at the time as of the START of the time
    step. Previously, CropPhenology looked at time as of the END of the
    time step. This was somewhat problematic, particularly because it
    meant that the last time step of the year was considered Jan 1, and
    so crops with a planting window beginning Jan 1 could be planted at
    the end of the previous year rather than the start of the new year.
    This was becoming particularly problematic in the context of Sam
    Rabin's upcoming prescribed sowing date work (see ESCOMP#1623 and ESCOMP#1616
    for some discussion).

(2) Call CropPhenology regardless of doalb (however, still do not call
    CropPhenology on time step 0). (See some discussion in ESCOMP#1626 and
    ESCOMP#1623.)

Rework single-point testing

Two changes to make the single-point tests in our test suite run more
smoothly. The main motivation for this is that single-point tests have
been failing frequently on cheyenne, presumably due to issues with nodes
in the share queue.

(1) Rework the test list to remove redundant single-point tests and move
    many of our single-point tests from cheyenne to izumi. (With (2) in
    place, this isn't really necessary, but given that we're going to be
    using full nodes for single-point tests, it seems better to do this
    on izumi rather than wasting all of these processors on the
    generally more heavily-loaded cheyenne. Also, this helps accomplish
    the goal of providing better balance between our different
    machine-compiler permutations, helping to reduce overall testing
    time.)

(2) Change run_sys_tests so that, for the remaining single-point tests
    on cheyenne, we use the regular queue rather than the share queue.
    This means that single-point tests use a full node, which is
    wasteful but makes the tests more reliable. (You can override this
    behavior with the --queue option to run_sys_tests: To revert to the
    earlier behavior, where the queue is determined automatically for
    each test, add '--queue unspecified' to your run_sys_tests
    invocation. (As before --queue can also be used to specify some
    other queue to use for all tests.))