Fix averaging for FH000 when using IAU

[dpsarmie]

Description

A bug was reported (NOAA-EMC/GFS#53) where the averaged fields in sfcf000.nc were not being calculated correctly. This occurred when in IAU mode and only affected the sfc file at fh000.

The fix involves adding a case for fh000 in the InitTimeFromIAUOffset subroutine in ./fv3/atmos_model.F90 where the correct averaging window is set. The functionality at other points in the model remains untouched. diag_time_fhzero was also initialized for non-IAU cases to prevent errors. This variable is not used in non-IAU cases so that functionality remains the same.

The variables that changed in sfcf000.nc in the GFSv17 tests were the following:

cduvb_ave
cpratb_ave
csdlf
csdsf
csulf
csulftoa
csusf
csusftoa
cwork_aveclm
dlwrf_ave
dswrf_ave
dswrf_avetoa
duvb_ave
evbs_ave
evcw_ave
gflux_ave
lhtfl_ave
nbdsf_ave
nddsf_ave
pah_ave
pevpr_ave
prateb_ave
pres_avehcb
pres_avehct
pres_avelcb
pres_avelct
pres_avemcb
pres_avemct
sbsno_ave
shtfl_ave
snohf
snowc_ave
tcdc_avebndcl
tcdc_aveclm
tcdc_avehcl
tcdc_avelcl
tcdc_avemcl
tmp_avehct
tmp_avelct
tmp_avemct
trans_ave
u-gwd_ave
uflx_ave
ulwrf_ave
ulwrf_avetoa
uswrf_ave
uswrf_avetoa
v-gwd_ave
vbdsf_ave
vddsf_ave
vflx_ave

No other variables or files at any other forecast hour were changed.

Issue(s) addressed

Closes #1080

Testing

This fix was tested using the GFSv17 operational configuration on WCOSS2. The fix was also tested using the regression test suite of the UFSWM on Ursa.