Ncurses declaration compatibility (52383 vs. 52398+)

[Bart Schaefer <schaefer@xxxxxxxxxxxxxxxx>]

We have competing patches for configure failures in the face of
 gcc -Werror=incompatible-pointer-types

Florian (52383) simply cast away the error in the configure script,
because the const-ness is inconsequential to Src/Modules/curses.c
usage of the arrays.

Nicholas (cf. 52398) did extra work for const correctness in the Src
code as well, but made some stylistic choices that Oliver and I found
superfluous or unhelpful.

To supply some context Nicholas did not have, users/13418 introduced
the "zstrcodes" variable because a Cygwin compiler in 2008 had issues
with (re)declaring "strcodes".  However, zstrcodes was described as a
"hack" that would have been better corrected by diagnosing the reason
configure choked on strcodes.  This was never done.  Aside from
Oliver's stylistic objection to the use of #defines, enshrining an old
hack by changing the handling of all the other ncurses arrays to match
it, seems a step in the wrong direction.

Any volunteers, with Cygwin access, to back out users/13418 and
determine if strcodes is still broken after 15 years?