Extract the V1 CSV recipient parsing and test it

[gyorgybalazsi]

batch::submit_from_csv parses its CSV rows inline, inside the
function. batch::v2::parse_recipients does the same work in a named
function, and only the V2 copy has tests.

So the CSV format rule is written twice, and the tests point at one
copy. A change to the format will edit v2::parse_recipients, because
that is what the tests name, and V1 will keep the old behaviour
silently.

Extract the V1 loop into parse_recipients beside its V2 twin, and
give it the same three tests: a good file, an unparsable amount, and an
empty file.