Some values that can never, ever, change were held in local
variables, rather than in PROGMEM. Fixed.
Change "pressed" to a signed int so the test for < 0 makes
sense, and to avoid possible weird failure modes in the
case where a key release comes in when pressed is already
zero. (Shouldn't happen, sure, but computers are weird.)
A lot of things in process_steno had external linkage for no
particular reason. They've been marked static. Stuff still
builds.
Distinguish between currently-held keys and keys that have
been held, and expose these values through a nicely-named API
so other code could, say, check on the current set of steno
chording in order to make displays. Also in passing fix up the
"state" value having external linkage so it could clash with
other people's variable declarations.
The API also provides hooks for key processing and steno chord
events, so you can monitor those events without having to
run in matrix_scan_user and recheck the values directly. Also
document these.
There is no path through processing a key that doesn't
end with a return false, so the nested return foo() are
gone and we just return false.
