parallel bug in wrfda readhdf5ahi thinning

[ivaNov-0042]

in WRF/var/da/da_radiance/da_read_obs_hdf5ahi.inc line 503
in root process, if num_ahi_global ==0 ( this happens frequently in some numproc settings, when domain not quite under sat sub-point ), then root process will not enter parallel thinning. but other process will still wait for root process reduce and broadcast the in(itxmax) and out(itxmax) array. This will cause entire wrfda stuck. this bug confuesd me a long time, while i'm trying to batch run wrfda for a month.