Use setpixels() for Sixels
Fix setpixels

It's WAY too slow... the average latency per function is set by the SLEEP(1)
on line 1641... I'm not sure we have a reliable cross-platform sleep more
granular, and 1ms is too slow if we're doing individual pixels.

cterm < 1.167.
Use supports_fonts() in xbin_draw() so we don't bother sending font data
when we know the terminal can't use it, but we still want to display
the xbin image data.

Make send_rectangle() static, nobody else should be calling this.
Clean up a bit... this is still badly broken though, especially clearing
for some reason.

If update_rect() fails, request a complete screen redraw
Move pixel redraw requests out to avoid full screen refreshes when possible
Comment update_rect() better
Consider a character updated of blink or bold is set, and the meaning of blink or bold has changed
Actually update a character if it's changed(!)

Since we don't use a separate lock, add a request_pixels_locked() function.

Hopefully fixes missed updates.

of the response escape sequence. This fixes a number of issues with "keys"
being interpretted (that weren't hit by the user) when activating fonts
and querying modes.
Also added "CTerm" to the debug log output lines.

A regular expression, but (I think) the leading/trailing slashes should
be omitted.
Files matching this pattern will not be listed.

at shared memory.

Cuts down CPU usage on my system from ~30% on an idle screen to ~3%.

(fixes problem in v3.16 or non-current v3.17 trying to load fonts into
non-font-capable terminals).

lock contention.

Unfortunately, there are no wrappers for pthread_rwlock_*(), so use mutexes
for Win32.

While we're here, clean up the locking regime a bit... in lock order, keep
the locks like this:

vstatlock
vmem_lock
screenlock

This should reduce future deadlocks (once I shake out any new deadlocks that
is).

detected screen mode, display a warning, but still display the file.
Many times, traditional .xb files (not xbimages) will look fine when
ignoring the embedded font data.

with the current detectec charheight and ignore any font data if they
dont' match (but still display the image data). This will sometimes result
in "garbage" on the screen, sometimes it'll look fine. Depends on how the
file was generated.

major *1000 + minor.
console.telnet_cmd() now accepts an optional 3rd argument, timeout (default 0)
which if non-zero will wait for an acknowledgement from the Telnet peer.

Over-hauled the ANSI detection a bit too.

Fixes sixels disappearing with selection issue.

function.  This allows simple fixing of clreol() and clrscrn() which were
not aware of 256-colour mode.

cputs and cputch aren't really needed anymore, but they may be useful, so
leave them in.

adjust the extended fg/bg after the call.

1) The colour values are expressed as a percentage, not a 0-255 value.
2) Implement CSI ? 80 h/l to set/reset the sixel scrolling mode (defaults to
   set).
3) Document sixel commands.
4) Combine the extended set/reset mode commands in the documentation now
   that multiple modes can be set or reset with a single sequence.

XBin (*.xb) files. The display dimensions of the files may be larger (wider and
longer) than the terminal window. You can use the arrow keys/home/end/num-pad
equivalent (including diagnals) to move the display window around.
If you have a recent SyncTERM 1.1b build, you'll get palette redefinitions
and (any version of SyncTERM 1.0 and later) you'll get the custom fonts too.
Note: the entire Bin/XBin file is loaded into JavaScript arrays/objects, so
very large files may require increaisng your JavaScript heap size considerably.

XBimages) and moved the read() method there.

this function for classic .xb files with custom palettes and support
the new SyncTERM/XTerm palette redefinition and reset esc sequences.

Darwin because it's easier, and FreeBSD so I can reproduce problems it
causes on Darwin.

With the supported compilers, we can't rely on the return value of snprintf()
and historically have replaced it with safe_snprintf().  For macOS however,
snprintf() is apparently a macro, not a function, so gendefs.h throws and
error due to redefinition.

Rather than hack around on that, just ensure that we never use the return
value of snprintf() and use safe_snprintf() to ensure we know what the
return value will be.