neumann boundaries calculated relative to border, not relative to
column/row. Shouldn't make a difference, but the latter leads to sag
faults probably due to vectorisation, as indexation executed
independent of conditional

shouldn't make a difference - not sure why it does

had to comment out one of the boundaries fixes, as it leads to a
segmentation fault right away. No idea why.

but boundary fix doesn't seem to work in this version - sets all fields
back to zero?

# Conflicts:
#	salt-marsh/src-v1/Main.hs

This is geared towards stack/docker integration, as a workaround for an IHaskell bug on macOS.

Still diverges after about 8k steps, but this looks like it's due to
the improper handling of the boundaries, so I'm leaving it as it is for
the moment.