publish QOS still defaults to 0 (for now at least), while
subscribe QOS defaults to 2.

And disable the mqtt log callback by default (it's very noisy)

Maybe help diagnose the disconnection/failure-to-reconnect issues I'm seeing.

Don't double-publish log messages when not in daemon-mode
Publishing to <host>/recycle will recycle all servers.
Log a successful broker connection upon startup.

Default MQTT version is 5
Default username, password are block. Up to 255 chars each.

Currently, sbbs.exe only
Includes non-Unix build fix in sbbscon.c for issue introduced in previous commit

Requires libmosquitto (e.g. install mosquitto-dev pkg) and set
in ctrl/main.ini:
[mqtt]
Enabled=true
broker_addr=127.0.0.1
broker_port=1883
qos=0
keepalive=10

Next up: Windows support

Needed a version of strListSplit() that doesn't treat consecutive delimiters (e.g. "\r\n\r\n") as a single delimiter: so introducing strListDivide(). Like strListSplit(), strListDivide() modifies the input string (replacing first occurrences of delimiter with NUL).

Since we don't want "\r\n" to be counted as 2 lines, we need to just split/divide on '\n' and then truncate any line endings (i.e. '\r') chars off the ends of the split strings.

Reported by phigan in #synchronet

Needed a version of strListSplit() that doesn't treat consecutive delimiters (e.g. "\r\n\r\n") as a single delimiter: so introducing strListDivide(). Like strListSplit(), strListDivide() modifies the input string (replacing first occurrences of delimiter with NUL).

Since we don't want "\r\n" to be counted as 2 lines, we need to just split/divide on '\n' and then truncate any line endings (i.e. '\r') chars off the ends of the split strings.

Reported by phigan in #synchronet

Replaces the status(), started() and terminated() callbacks, in this case.

set_state() is called from a single thread (unlike status() was), so no mutex should be needed here.

Don't do the cursor dance. We have multiple servers, so reflecting their state in the cursor is not really accurate.

I'll clean-up the commented-out code after some more testing.

When the systemd dev package (e.g. libsystemd-dev) is installed, we can use
sd_notify() to inform systemd of Synchronet server state/status. The state
values (READY, STOPPING, RELOADING) are "well known" while the status
value is a free-form text string (the last lputs output).

Eliminated the free-form status() callback from *startup_t.
Eliminated the now-redundant started() callback from *startup_t.

Eliminated the use of sbbs_status.c/h as nothing is using that interface
and while I do plan on using a remote control/monitor interface, I'm
considering use of a standard pub/sub lib and protocol.

This commit is going to break the Windows sbbsctrl build for sure.
Probably break the sbbsNTsvcs build.

New LINEDELAY and LINEDELAY:n @-codes
Another way to pace the terminal output of long display (e.g. ANSI) files. LINEDELAY enables a 1/10th of a second delay before the sending of all LF chars. LINEDELAY:n sets the line_delay duration explicitly to a value in 1/100ths of a second units. So LINEDELAY and LINEDELAY:10 are equivalent. LINEDELAY:0 turns line-pacing off. The original line pacing value is always restored after a printfile/putmsg operation.

New JS console.line_delay property (milliseconds) to control via JS. Setting this to a really high value would be bad. Some range enforcement should be added to this and many other console control values (TODO).

This addresses issue #459 (Synchronet never allowed SSH-auth via real name)

Also, localize the "valid real name" checking logic into a single new function:
check_realname()

No longer applying QWK filters to incoming netmail messages since we would need different criteria for netmail filtering (e.g. don't apply the twitlist.cfg). And I'm not clear that we actually need netmail filtering.

Disable the importing of "remaining" 
VOTING.DAT sections (polls, votes) since this should no longer be necessary for backwards compatibility and appears to the source of errors like:
qwk.cpp line 1141 (qwk_vote) writing "/sbbs/data/subs/sync_sys" access=-105 info=smb_addpollclosure thread_back field missing

qwk_vote() takes a msg_filters argument, so we'd have to pass that to qwk_handle_remaining_votes() and frankly, it just shouldn't be needed any more, so disabling and logging a warning if there are any "remaining QWK voting.dat sections" after parsing QWK/REP packet (which consumes the sections from the parsed VOTING.DAT file).

I just happened to notice qwknode* in sbbs.h and was like: what is that? Oh, it's a realloced array created when parsing QWK routes. When is it freed? Oh, never.

This bug was introduced in commit 9c91d32a 2 years ago and would result
in votes and poll message being erroneously filtered. e.g.
!Filtering QWK message from (null) due to age: 19321 days

The filter criteria was either a zero-initialized msg or the previously
parsed/imported msg from the QWK or REP packet.

A solution for filtering vote/poll messages is needed, but this isn't it.

This reverts commit 2cddddce.

This change was "wrong-headed" per Deuce and broke JS sends on
blocking sockets. Thanks for the help.

This fix for inifinite-wait on send() likely led to the infinite
BinkIT errors/log messages that led to commit b795cf6a.

The real fix for this problem (which can block all other timed
events from running), would be a Socket.poll() on the socket before
send in binkp.js.

command -> command_shell (reported by Nelgin)
expire -> expiration_days

More obvious (the change) when running with a bigger terminal.

Faster as the sections have already been parsed.

js_open() sets the rval for the context (e.g. to JSVAL_TRUE), so we need to set
it back to JSVAL_FALSE for failure conditions. Otherwise, the methods always
return true even upon (save-msg) failure.

If nodeX/node.cnf has the wrong node number, the [R] flag would be cleared from
the wrong node upon automatic configuration reload. This likely fixes the issue that Keyop is reporting in #synchronet

This allows the sysop to configure the settings to be used for newly-created and auto-added directories without requiring any "template" directory to be configured. The "template" directory concept goes away (so in this way, File libs/dirs are now different from Message groups/subs in how bulk configuration is done).

The Lib->Clone Options menu option was moved to Lib->Directory Defaults->Clone Settings and no longer uses the first sub or the template sub but rather the configured directory defaults to be cloned into all dirs in the lib.

I decided to make the dir_defaults their own sections in file.ini rather than combining with all the other lib settings (and deconflicting key names such as "sort" and "settings"). This also allowed me to use the same cfg read/write functions for those directory settings and lib's directory default settings.

The configured path could be absolute or relative (to the parent). The trailing (back)slash is important to the string comparisons.

Placing a colon in a group's or lib's short name would prevent any sub-boards within that group (or directories within that lib) to now be parsed since we were truncating at the first colon (rather than the last) after the section prefix.

An internal code is now longer than a group/lib short name!

Upload and Download ARS must be matched *and* the directory's upload and download ARS.

Operator and Exemption ARS must be matched *or* the directory's operator and exemption ARS.