When SyncTERM started by enabling binary mode in both directions,
the internal tracking of the binary status wasn't updated, so it
was incorrectly tracked as being in NVT mode.  After a file
transfer, it would then revert to the NVT mode it throught it was
in.

This change updates the binary mode value when sending as well as
receiving TELNET_BINARY_TX.

This is still technically broken though since binary mode is
negotiated separately in each direction, and the initial send of
WILL + DO is actually a pair of requests that need to be confirmed
by the remote.  Until they are confirmed, the connection is still
in NVT mode.

Hopefully though this isn't an issue since the remote should reply
to both, and if it denies there's no effective difference between
what we should do when already in binary mode and not because we
don't support any other modes (such as CHARSET option).  Fixing it
correctly would get very complex and involve blocking the connection
until we get a response.

as pointed out by Nightfox.

... just the returned copy (when toggling the DST flag).

This resolve the issue of SCFG->System menu detecting changes by the sysop
(and prompting to Save Changes when exiting) when no changes were actually
made, just that DST had been auto-enabled/disabled based on the date.

rlogin_xtrn_logon set to false to skip the normal logon screens for Rlogin.
rlogin_xtrn_logoff set to false (default) to immediately hangup, true to
prompt the user if they truly wish to logoff, and "full" to prompt and if
affirmed, go through the full logoff procedure (dislpay files and events,
etc.).

Otherwise there's a race condition that will crash Win32 at least.
Fixes ticket 199.

This crashes SyncTERM on Win32 if it's the first ANSI music played

Also, include WaveOut in the audio build options.

As requested via DOVE-Net by xbit and StingRay

If the sysop wants custom text.dat strings to be unique per terminal
connection protocol (e.g. telnet, rlogin, ssh, raw), this is how that
can be achieved (e.g. ctrl/rlogin/text.ini can be used to override text.dat
strings only for RLogin connections).

To at least partially address issue #897

Really, a single source of truth for the version major.minor number now.

Use math to compute the value of VERSION_NUM and VERSION_HEX so we don't have
mutliple sources of truth for the integer-encoded version number any longer.

Makes it nicer to read through looking for specific commands

These all have slightly different semantics (that don't matter yet),
so split the up appropriately.

I really shouldn't be doing this though. :(

This allows for a way to save/restore the cursor position when
DECLRMM is enabled.

Saves to the same variables as SCOSC/SCORC.

Document that ANSI.SYS is what's implemented, and indicate how this
can be extended.

Closes issue #896

These are zero-based like the properties, not 1-based like gotoxy()

We only do this in the terminal server... to detect this, check if the
global object has a bbs object that is of the BBS class.  If it does,
the context private will be an sbbs_t pointer.

For now, this is a quick hacky fix.  The web server does not store
an sbbs_t in the context private, so we can't use it here.

This needs further work.

Initial design of Terminal class

See merge request !506

Better/more obvious "-1" bytes/size logged when flength() returns failure.

This broke showmsgavatar.js

Examples:
"-D02/29/25"
"-T02 Feb 1996 03:04:05 GMT"
"-TDecember 17, 1995 03:24:00"
"-T1995-12-17T03:24:00"

The system's configured time zone is always used however.