integrate the preprocessor for forcings

Which NextGen forcings processing algorithm is the quickest can depend on the size of the requested processing job in terms of number of catchments and number of time steps. The currently integrated forcingprocessor is optimized for NRDS forecasting configurations of VPU scale and relatively few time steps (<250). The preprocessor is better suited for retrospective configurations of ~100 catchments and thousands of time steps.

Datastreamcli could contain logic to decide which integrated tool to use. The preprocessor is now planned to be integrated as a subsetting and bmi config provider, so it'll be available for forcings as well. The cleanest path to this could be to make the choice inside of configure_datastream.py. and have the datastream script run forcingprocessor if that config file is present, or the preprocessor if not.

Start in configure_datastream.py and inspect how the datastream script currently selects and runs forcingprocessor. Check how the preprocessor will be exposed as a subsetting and BMI config provider, then implement and verify the forcings selection between forcingprocessor and the preprocessor based on the available configuration.

Written by the indexing model from the issue text.