Just rename ESCDELAY member to escdelay.

Apparently, the ncurses in openSUSE Leap-15.2 has the following:
Since we have a struct member named ESCDELAY, we can't actually set
it at all when this macro is defined.

The three possibilities to deal with this are:
1) Rename the ESCDELAY member.
   Likely the best option long-term, but I don't want to find everything
   and update it at this time.
2) Avoid using the ESCDELAY member #ifdef ESCDELAY
   Which makes it not work in the exact set of circumstances it makes
   sense in.
3) #include curs_fix.h from ciolib.h
   This actually renames the ESCDELAY member to _nc_ESCDELAY, but you
   won't be able to actually notice this.

1) Ciolib CIO_KEY_QUIT and CIO_KEY_MOUSE conflicted with ALT_7 and ALT_6
   respectively, change these to use \xE0 prefixes.
2) Add a new dorkit key "CONNECTION_CLOSED" which is returned by getkey()
   after the connection is closed.
3) Have the various connection back-ends send a CONNECTION_CLOSED key to
   the keyboard buffer when the connection is closed.
4) Have the local console send a CONNECTION_CLOSED key when the window
   is closed.
5) Disable auto_terminate in dorkit.
6) Introduce dk.connection.disconnect_timeout variable (defaults to 30
   seconds).  After a CONNECTION_CLOSED key is received, dorkit will
   immediately set js.terminated, then after disconnect_timeout seconds,
   will enable js.auto_terminate.
7) After a CONNECTION_CLOSED key, waitkey() always returns true, getkey()
   always returns a CONNECTION_CLOSED key, and getbyte() always returns
   undefined.
8) To prevent a poorly written door from hogging 100% CPU for the whole
   30 seconds, put an mswait(1) in the connection check when the
   inactivity timer is running.

There are a number of conio changes in here too... too many to list.  Check
the diffs if you actually want to know what changed.

These allow setting the initial window size for modes that don't force
integer scaling (ie: SDL).

While we're here, clear a renderer before free()ing the texture to prevent
crashing on some nVidia cards.

Addresses https://sf.net/p/syncterm/feature-requests/17/

There's still some shittiness with the keypad, but nobody really knows how
that works anyway.

Add support for scroll "buttons" (buttons 4 and 5).

Remove xlat.
Rearrange Commdore fonts to match ASCII.
Clean up C64 emulation a lot...
Set the Commodore background to black in CTerm... it seems pretty much
all terminals do this.

getcliptext() and copytext() (ie: copy/paste) now work on UTF-8 strings.
Not tested on Win32 yet.

Allow setting a '0' for left/right margins to leave them unchanged.

compatible (most terms are these days).  In this case, curses mode works
correctly.

Add a new CIOLIB_MODE_CURSES_ASCII for the old behaviour.

Add new ciolib_appname and call SDL_RegisterApp(ciolib_appname, 0, NULL)
when the symbols are loaded.

Have SyncTERM save/restore the window size

Ciolib changes:
Hopefully fix streaming texture updates (ie: white screen)
- Request pixel data on expose rather than re-render texture
- This allows the rect update to copy the screen
Fix ALT-Enter fullscreen toggle
Ensure vstatlock is held when accessing cvstat
Add setwinsize() and setwinpos() to ciolib
Return window position from get_window_info() in SDL mode
SDL driver no longer supports setscaling() (use setwinsize())
Use ALT-<arrow> to change window scaling, not Meta
Add winwidth/winheight to vstat structure

Remove CIOLIB_MODE_SDL_YUV* macros, these modes have no point anymore
Fix SDL prototypes

Use _beginthread() instead of SDL_CreateThread().  For some reason, using
CreateThread() sometimes caused an application crash (ie: when using -?)
Everyhing else uses _beginthread() so there's no reason for this to be
special, especially when it breaks stuff.

Win32 now (and it only runs main() from WinMain()).  This shouldn't be
necessary for MSVC builds.

It's not unlikely that any utilities using SDL will have new issues... you
may want to hold off on updates for a bit until this settles unless you
don't use SDL mode.

vidparam structure itself.

a struct vmem_cell.

Also, the vmem is an array of struct vmem_cell.

Fix age-old bug with Borland/C++Builder built executables (Windows):
to achieve compatibility with  the default __cdecl symbol naming rules of
Visual C++, we were using __stdcall convention for DLL functions when
building code with Borland/C++Builder tools and using the default (__cdecl)
convention when building with Microsoft (Visual C++) tools. Although this
allowed symbols to be located when linking, the calling convention mismatch
caused a stack cleanup issue that very rarely manifested itself in a bug
(e.g. exception of some kind in sbbsctrl.exe, usually). Mismatching
the calling conventions was unintentional (I thought the default for MSVC
DLL functions was __stdcall) - but since the calls to MSVC-Built DLL functions
worked 99% of the time, I didn't realize there was an underlying issue. So I
now work-around the DLL symbol naming mismatch using a command-line option (-a)
passed to implib in src/sbbs3/ctrl/makelibs.bat

I had previously worked-around exceptions when calling MSVC DLL functions in
sbbsctrl.exe by calling the problematic DLL functions from a timer tick handler
rather than a user control (e.g. button) event handler. Those work-arounds can
now be removed.

The erroneous "DLLCALL" definition design pattern was replicated (copy/pasted)
to many other projects' header files in cvs.synchro.net. In the future, we may
want to just remove all instances of *CALL since they now serve no purpose and
appear as useless "Kruft" (but do allow us to more-easily globally change DLL
function calling conventions if/when necessary in the future).

current mode.

Font 0 is always supported in all modes.

The biggest impact is in the cterm =1n query, which is retaining 0 for
success and now uses 1 for all failures.

Update scrollback to use vmem_gettext()
Update all the things I previously updated for the new cterm_init() prototype

So now I expect things to work again... though X11 mode may have issues,
and SDL overlay mode is dog slow.

We're switching to TrueColor now, and maintaining a 65536 entry palette
for the various palettized bits.  This change hurts the overlay output most
since it currently means floating point calculations for every pixel.

The getfont() function now takes a fontnum argument so you can get all the
current fonts.

The vmem array now holds the font data, so once a glyph is written to the
screen, it can stay in that font "forever".  This still has some issues
with savescreen()/restorescreen() since there's no extension to grab the new
stuff... that'll be coming soon, likely replacing the pgettext() and
pputtext() functions.  For now, various popups will likely reset fonts to
the default font on the screen.

Now that we have all this, the status bar in SyncTERM can retain the default
font and colours at all times.  Further, menus can be displayed in the "right"
colours and fonts, and no more translation will be needed.  This is a work
in progress though, so there's likely to be glitches.  The scrollback doesn't
have the font data either (same issue as savescreen()/restorescreen()).

1) XTerm using semicolons.  CSI 38/48 ; 2 ; 0 ; R ; G ; B m
2) PabloDraw/TundraDraw. CSI 0/1 ; R ; G ; B m

These use a temporary internal palette of 15840 entries to ensure what's
on the screen is correct.  Scrollback data may be incorrect however.

This hack is still wrong, but at least it won't break builds that don't use
bitmap_con.c

These are 132 column modes which use 8x16 fonts and have 16:9 and 5:4
aspect ratios respectively... DigitalMan hates them!  ;-)