Not tested on Windows.

Basic Prestel support based on the "Prestel Terminal Specification"
Issue 2 by the Post Office.

Known issues:
Modifying an existing line does not update the row after it.
Double height is not supported.
Keyboard mapping kinda sucks (looking at you #/Return).
Concealed is always shown.
Remote programming sequences aren't supported.

There appears to be something wrong with the End of the Line title
and menu screens, but maybe nelgin did them wrong.

Now that the conio layer is handling translation of character sets,
the key handler in term.c should only handle terminal related
translations (such as backspace to delete).

This fixes bug 159, but there's still some inconsistency across
the various backends on how they handle C0 control characters and
input characters outside of US-ASCII.

Windows appears to treat backslashes and slashes the same when
using the API, but Windows always returns backslashes.  Convert
backslashes to slashes in clean_path() to ensure paths compare the
same regardless of them being specified with backslashes or slashes.

There's still an open question of what to do about paths received
from the remote with backslashes in them, but I suspect that doing
that would actually work on *nix systems and break on Windows systems,
so hopefully whoever does that can fogure out what's going on.

Using snprintf() prevents the last character of the date from
being put into the string.

RIP is currently the only thing that can change the number of lines
in the terminal while connected.  This can actually result in there
being more lines on the screen than the last set video mode supports.

There's a "feature" in SyncTERM text modes (curses, ANSI, and Win32
Console) where if the screen mode is "current", it uses the number
of lines in the terminal, but if it's anything else, it forces the
number of lines to the lines in the mode.

This works great until RIP increases the number of lines beyond the
number in the mode... when this happened, RIP would draw the status
bar at the incorrect location on the screen.

This change no longer sets the max rows in the terminal if the
backend can set individual pixels (required for RIP).  Further, to
avoid a delay before redrawing the status bar after RIP changes the
number of rows, the reinit_screen() function now sets a global to
indicate it was called, and the status bar will always redraw when
it's enabled and the flag is true.

This *may* be related to bug 140, but may not be.

It's likely time to find a nice format that can be used to generate
HTML and PDFs and such now.

The option from the menu (ALT-Z/CTRL-S) was copy/pasted years ago
and never updated while the keyboard shortcut (ALT-L) was updated
quite a bit.

This extracts the ALT-L code into a function and calls it from the
menu.

This allows easy copy/paste without scrollback hacks on BBSs that
insist on taking over the mouse just so you can move the mouse and
click a character rather than press one key because it's 3l337..

This allows ALT-B from dialing directory to show the last seen
thing, not the extra blank screen added to the end.

Doesn't fix anything, but a huge QoL improvement for people who use
ALT-B after disconnecting.

With 0xE0 being used for ciolib "super-extended scancodes", a literal
0xe0 can't pass through the input path.  This is an issue in CP866 (р)
and KOI8-U (Ю) as well as CP437 α.

Should fix SyncTERM SF bug 123.

As reported by GCC 10.2 with FORTIFY=1 and SANITIZE=1

- Support LCF (Last Column Flag) mode per DEC STD-070
  (Despite years of protest that it's stupid)
- "Properly" support CSI 7 m and CSI 27 m
  Now, when in "Negative Image" mode, changes to the foreground
  change how the background is drawn and vice-versa.  Perviously
  this command just swapped the two values and called it good.

Fix bad and poor unicode mappings, add support for sending Insert
and Pipe.

Toggled using backtick.
Current state is shown in the status bar.

Since we previously couldn't actually *show* the status bar in
ATASCII (or C64) mode, enable that as well.  This means supporting
40-column status bars too.

We still default status bar to off for those modes, because that's
likely what people actually want.

Also, it seems ENTER was broken in ATASCII mode, so fix that as well.
Implements SF feature 61

warning: ‘maxval’ may be used uninitialized
warning: ‘inch’ may be used uninitialized

At the same time, add BD, BE, PE, and PS to the terminfo entry.

Note that it seems this is "normally" detected by seeing if $TERM
contains "xterm" which shouldn't work with SyncTERM (which is wildly
incompatible), but the terminfo source file here:
https://invisible-island.net/ncurses/terminfo.ti.html
Gives us hope in the form of this comment:

https://invisible-island.net/xterm/xterm-paste64.html

Bracketed paste was introduced by xterm patch #203 in May 2005, as part of a
larger feature for manipulating the clipboard selection.  Few terminals aside
from xterm fully implement the clipboard feature, but several copy this
detail.  The names for the extended capabilities here were introduced by vim
in January 2017, but used internally.  In 2023, vim patch 9.0.1117 is needed
to work with this change.

That is to say that it likely won't work on anyone's system today
(except maybe Cyan's), but it may magically start working in the
future... assuming tic supports these capnames.  No real clue there
since there's absolutely no termcap support, and I use FreeBSD.

Some Linux systems ship with a reentrant ncurses library that does
not have an ESCDELAY variable.  For now, simply don't allow control
over ESCDELAY on the systems since we also currently have a global,
and so can't use set_escdelay() as intended.

Also, in ripper, stub out all the stuff that's used when HAS_VSTAT
is not defined.  This makes most of the code not compiled when it's
useless anyway.

Fix a couple warnings while we're here.

warning: format ‘%s’ expects argument of type ‘char *’, but argument 4 has type ‘void *’
warning: ‘sprintf’ writing a terminating nul past the end of the destination