FATES hydro mass-balance issues

### Brief summary of bug

FATES-hydro is a special mode where vegetation can uptake and store water via the transpiration stream. This requires an accounting of how much water is stored in FATES vegetation, which must be passed back to CLM for its water mass balance checking. The water mass balance check (BalanceCheckMod.F90 at line 760) is failing. Details below.

### General bug information

**CTSM version you are using:** [output of `git describe`]

hash: 3ac4305
tag: ctsm5.1.dev029-3-g3ac43057

**Does this bug cause significantly incorrect results in the model's science?** [Yes / No]

Not sure, but I don't think so.

**Configurations affected:** [Fill this in if known.]

These namelist options must be set:

use_fates = .true.
use_fates_planthydro = .true.

### Details of bug

I generate this error using a specialized site setup for Barro Colorado Island Panama, where I also initialize a forest stand structure from inventory data. The simulation runs 26+ months and then fails with the following error:

```
clm: calling FATES model 38449
FATES Dynamics: 1902-03-13
dbh counter: 13 is woody: T
dbh counter: 13 is woody: T
dbh counter: 17 is woody: T
dbh counter: 17 is woody: T
clm: leaving fates model 1 1
WARNING: grid cell-level water balance error nstep= 38449 local indexg= 1 errh2o_grc= -2.2061354929889532E-003
CTSM is stopping because errh2o > 1.0000000000000001E-005 mm
nstep = 38449
errh2o_grc = -2.2061354929889532E-003
forc_rain = 0.0000000000000000
forc_snow = 0.0000000000000000
endwb_grc = 2989.3396629148806
begwb_grc = 2989.3595745851790
qflx_evap_tot = 1.9854765757965773E-002
qflx_sfc_irrig = 0.0000000000000000
qflx_surf = 0.0000000000000000
qflx_qrgwl = -2.1492309525845731E-003
qflx_drain = 0.0000000000000000
qflx_ice_runoff = 0.0000000000000000
qflx_snwcp_discarded_ice = 0.0000000000000000
qflx_snwcp_discarded_liq = 0.0000000000000000
deltawb = -1.9911670298370154E-002
deltawb/dtime = -1.1062039054650086E-005
qflx_drain_perched = 0.0000000000000000
forc_flood = -0.0000000000000000
qflx_glcice_dyn_water_flux = 0.0000000000000000
CTSM is stopping
local gridcell index = 1
global gridcell index = 1
gridcell longitude = 280.15390000000002
gridcell latitude = 9.1530000000000005
ENDRUN:
ERROR: ERROR in /raid1/rgknox/SyncLRC/ctsm/src/biogeophys/BalanceCheckMod.F90 at line 760
```

From the log, I can see that the check is failing after a FATES dynamics call. I have to investigate further to see a call to FATES hydraulics follows. If only the dynamics call has transpired (pun intended) prior to the balance check fail, then we should expect that that the total water stored in vegetation should had remained unchanged during the process. During dynamics, we only shuffle water around internally in FATES, but should conserve it. We do have a check in place to verify this (but checks can be wrong too of course).

Key variables that FATES passes to CLM are:

The total water stored in vegetation, an example of the transfer found here:
https://github.com/ESCOMP/CTSM/blob/master/src/utils/clmfates_interfaceMod.F90#L1011-L1012

The water uptake by roots:
https://github.com/ESCOMP/CTSM/blob/master/src/utils/clmfates_interfaceMod.F90#L2632

Key variable that CLM passes to FATES:
Transpiration rate of the patch:
https://github.com/ESCOMP/CTSM/blob/master/src/utils/clmfates_interfaceMod.F90#L2752

### Important details of your setup / configuration so we can reproduce the bug

I will need to reproduce this error on a supported resolution, this error was generated on a user-defined single cell grid.

Here is the compset and key namelist variables though:
2000_DATM%CRUv7_CLM50%FATES_SICE_SOCN_SROF_SGLC_SWAV
use_fates = .true.
use_fates_planthydro = .true.