getnodedat() now performs a non-locked read by default.

Perform more checks of getnodedat() return value before calling putnodedat()
to avoid unintentionally zeroing out node.dab records.

Add/use unlocknodedat() method for unlocking a node.dab record without
writing.

Note: The userdat.c getnodedat() and putnodedat() still return int (i.e. 0 on
success), so that can be confusing.

It's possible now for clients to easily exceed the configured max concurrent
connections limit, even though they just get an error 429 page. Let's at least
track and log when a new highwater mark is reached.

Can't backup (make copies of) an open file.

We can't send a 4xx error (ie: client did something wrong) in
plaintext on a TLS connection.

Untested beyond a basic compile check.

Clean-up the web server config menu quite a bit by condensing options.

You can also disable filebase access by setting the vpath prefix to blank.

We need to negate the negative return value to get the number of errors.

We have this feature for the FTP, Mail, and Terminal servers (with no default
limit) but crazy abuse of my Synchronet web servers has now made this a
necessity for Vertrauen. A limit of 2 used to be considered reasonable long
ago. Today, 10 seems fair enough. Setting to 0 will impose no max-connections
per IP address (other than the total client limit per server).

Similar problem as PulseAudio, but even grosser.

The macro was only define in the xpdev GNUmakefile, not in Common,
so bbslist.c did not know if it was enabled or not, and always
showed it as disabled.

Apparently JSVAL_IS_STRING() can return true even when argv[1] is out of range
(I guess I assumed it'd alwasy be NULL or VOID/undefined) and this can lead to
crash when calling JS_ValueToString() on it.

As seen on MacOS, an SMB shared file can cause the Mac SMB client to report EIO
error on locked-file read or lock attempts. So retry.

This makes more sense.  Also, ones that are meaningless are hidden
now.

Use 0 for "no held mosaic" to avoid confusion around printable bytes.
Store the state of the current cell *before* updating in fix_line()

Top/bottom detection was somewhat broken.
Interestingly, now the previous commit isn't working anymore.
So issue 172 is not quite fixed yet.

This was due to two main bugs... the simple one was a type where
the cutoff between separated and contiguous mosaics was incorrect,
which caused the first four separated mosaics to not be displayed.

The second was due to a number of bugs in the Hold Mosaic
implementation.

This fixes the first half of issue 172.

We want ASCII to map directly, and unicode values need to be sorted
ascending for bsearch() to work properly.

Fixes Prestel keyboard input not working in GDI mode.

If the remote sends 8-bit C1 codes, handle them as though they were
properly escaped.  Since Prestel terminals are 7-bit, these will
never be sent by a proper Pretel system, but apparently some systems
do weird non-standard things.

to notify sysop (user #1) that they were paged. Previously, this was
a hard-coded string. The new string in text.dat is the same, but adds
2 BELs (CTRL-G chars) and is now customizable.

Created an sbbs_t::notify() method to make this easier to use and that
method supports @-code expansion, so this text.dat string supports @-codes
(which I used).

This addresses the feature request in issue #828 (beep the sysop in the
telegram/short message, when they're paged).

1) Initialize saverobot.name since this is checked unconditionally.
   This has a side effect of initializing everything else, but name
   is the only thing that actually matters.
2) Add a bar to the global settings menu.
   Previously, if you scrolled near the bottom and selected an option,
   after changing the option, you would jump up to the last item on
   the initial screen.
3) Fix the AreaMgr Support default value to be the current, not the
   opposite of current.
4) Use the correct variable for EchoMail Relay Filtered Messages
   setting default.

We're carefully counting the errors, but never looking at the count.

For retina lies.

Upon any node.dab lock or read failure, this code would cause errormsg() which
would often/usually end up claling getnodedat() which would block forever trying
to acquire the ndoefile_mutex (introduced in commit b9633069, I'm not clear why).

Unlock/release the mutex *before* calling errormsg().

Don't deref that null pointer

It was always displayed (sent to stderr) when running scfg, echocfg, uedit,
etc. in auto-curses mode, so suppress that noise.

Some USB modems (reportedly, USRobotics USB modem) don't strip the parity bit
of data received from modem connections operating in < 8 bit modes
(e.g. 7-E-1), as is normally don't with a modem connected to a UART, so we'll
do that stripping (forcing to 0) here, as recommended by Deuce.

Nelgin, if you're doing some manually stripping of the 7th bit bytes received
from 7-E-1 connected modems and then sent to your server/BBS over TCP, you
shouldn't need to do that now.