DDLightbarMenu: Remove the inputTimeoutMS property; use console.inactivity_hangup instead.

See merge request !492

This reverts commit abdeb48f.

Apparently the Clang on the x86 Mac mini build node doesn't support this option

Ensure the stream is in a meaningful place.
Fixes a bug, but one we're unlikely to have ever seen the way we use
this.

The SyncTERM cache uses the BBS name as the directory name, so any
system-level caching needs to either be in the cache root directory
or in a directory whose name is guaranteed to not collide with an
entry name.

This leaves two options... either prevent a list of suffixes from
being used on BBS names, or have a separate system cache whose name
cannot be used as a BBS name.

I've taken the second route here... SyncTERM will not load a BBS
entry with the (case insensitive) name "syncterm-system-cache" and
will not allow creating one.  This is the directory where "global"
cache files will be stored (ie: scripts, internet dialing
directories, etc.)

It looks like at least one bus error was when the linker was loading
the .so.  Maybe a sync(8) here will "fix" that?

This is in the fastpath, but it's on the slow side where I don't
mind the occasional extra silly check.

Apparently, Windows doesn't have a MSG_DONTWAIT flag to recv().

In support of feature request 86.
Now, there's no immediate plan for what 86 is actually asking for,
but being able to point to things like the Synchronet BBS List and
having them automatically show up in your dialing directory would
be pretty nice.

Actually doing 86 means being able to merge lists, which I'm not
sure I want to deal with at this time.

(file cabinets, folders)

Bold the directory/library names and other markup improvements.

Expand viewable types to more readable archive types supported by libarchive
(many more than the list in Archive.supported_formats).

Remove an apparent nnecessary reset of console.aborted flag

Also, mention that SDL2 is required for macOS and Haiku

Regression introduced in 47b6f7a7

It seems there is a "hold mosaic" mode, but it doesn't completely
avoid look-behind when it's set.  The rules for what "clears" the
held mosaic when not in hold mosic mode are a bit murky... added
extra2.pst to test.

And merge the single/double height ones into the same function.
Normalize all the types while I'm here.

With this, the happy path draw_char_row_fast() is very simple and
easy to keep updated, and all of the complexity goes into
draw_char_row_slow().

There seems to be around a 20% performance difference between them,
but it can be hard to tell for sure because of the time spent
cheating.

Note that it doesn't appear to be worth cheating if we're going fast,
but that's getting a bit too deep for my tastes.

This has only ever worked briefly, and makes the code more confusing.

Also, change the terminal type reported to "Beeb7"

This is almost exactly like Prestel, but it's for BBS Micro BBSs
such as BeeBS.  The basic changes are:
1) Moving off the bottom of the screen scrolls instead of wraps
2) BBC Micro "high ASCII" is supported
3) Return sends \r, not #

To support the high ASCII, I actually implemented the ITU-T Rec.T0101
G1 set, which means the smooth mosaic characters are theoretically
available.

However, the SAA5050 emulation clobbers this by "correctly" turning
them into "blast-through" alphanumeric characters.

Doing this required the separated mosaic support be moved into the
font rendering and out of the font itself.

Anyway, the Sysop of beebs.ddns.net should be happy... and since
this will advertise a terminal type of "BBC Micro", they should be
able to auto-detect between ANSI and BBC Micro during login so the
user doesn't need to enter it.  Hit me up on IRC if you want help
with that.

Indenting lib and dir lists use the ul tag, I'm sure this isn't technically the
most correct way to do this, but its providing the cosmetic result I want.

Should make DigitalMan happy by making help output be in the console
on Windows with syncterm.com.  Hopefully syncterm.exe doesn't get
the help output hosed by this.

Also, if the output mode is ANSI, don't require a key press to exit.

This means the crappy 40-column status line can be used!

Assuming I'm right that is... if I'm wrong, this is merely what they
look like in SyncTERM today.