Speed up data processing via use of 'cfp'

The `gldas_get_data.sh` script processes its data serially by time period.  This can be optimized by processing all time periods in parallel using 'cfp' on Dell and Cray and its equivalent on Hera.