the history design here is… “interesting”

this prevents a segfault on “Esc - eof," " Return” on a file longer
than the screen, or on the new ^K] on it… and the value was only used
outside of macro execution in the first place already ANYWAY…

offenders will be patched away

• PARAMS begone
• libc4 ustat definition: resolve inline
• ignore HAVE_*_H for headers mksh includes unconditionally

• the joe 3.x/4.x/jupp bug only applied to files whose last nōn-newline
  character was Unicode (ipv ASCII), revert
• redo to correct that
• fix wsp-at-EOF trimming files that, after wsp-at-EOL trimming, are
  comprised of only newlines

This series of fixes is sponsored by ⮡ tarent

- empty file
- file with exactly one LF at EOF (DOS: also CR+LF)
- file with multiple LF (DOS: also CR+LF) at EOF
- file without any CR or LF at EOF
- absence of a RE bug in JOE 2.8 that all later versions have
- different RE syntax in JOE 4.1+

I’ve not tested CRLF with ≠2.8 but it “ought to be fine”.

Also, move ^K] and ^Q= into :def blocks and rearrange slightly.