For a project this small size with a small type ontology, it's better
C style to have one header where all the assumptions about types and
structures are immediately visible.
This is a first step towards turning all that global state into a handful
of structures - probably just two, one saveable as a binary game state.
No logic changes. A bunch of globals turn into fields (with the same names
except for lowercasing) in struct game_t.
Eventually this will allow drastic simplification of the save/load logic.
The all-capsing was a FORTRAN remnant.Also, we change a few FORTRANisms
so they are less confusing in this C context; ".TRUE." and ".FALSE." become
"true" and "false", "MOD" is mapped to % in places tha t are like C expressions
and (usually) "modulo" in places that aren't.
