Adjusting to account for slowness of writes for output when output
dataset directory is actually a mounted object store bucket (the default
at the moment), such that we work in container-specific dirs and archive
things before writing to such a place; also adding TODO comments to come
back and add more general behavior later, when output datasets may be
somewhere else.

Renaming NGEN_OUTPUT to NGEN_CSV_OUTPUT, introducing
ARCHIVED_NGEN_CSV_OUTPUT format for when CSVs are bundled into tar file
before being placed in dataset, and proactively add NGEN_NETCDF_OUTPUT
format for eventual use; also, switch to ARCHIVED_NGEN_CSV_OUTPUT as
current job default.

Bumping versions of dmod-core and dmod-communication for direct changes,
and then bumping other library and service packages to use the updated
versions for their dependencies (along with bumping their own versions).

Switching to use this instead of ValueError if command arg doesn't match
something supported by main(), and making message reflect the situation
a little better.