corresponding *_net_addr field is present (not NULL).

Don't strip the "user@" portion of a *NETADDR header field if the specified net-type is NET_NONE.

Allow an added or modified message header to have to/from/reply-to "net type" values of NET_NONE. This is a weird scenario, but it is technically possible (e.g. for nntpservice.js) to save a message with a to/from/treply-to "net address" header, but an associated net-type of NET_NONE. By saving the net-type NET_NONE, when saving a modified header, if the associated net address header field value cannot be parsed into a valid network address, there won't be any error reported, e.g.
Error -110 adding SENDERNETADDR field to message header

warning: iteration 9 invokes undefined behavior

events.

Automatically detect the character set of QWK-imported messages (that don't already have a FidoNet "CHRS" header) and create/set the FIDOCHARS header field accordingly (UTF-8, ASCII, or CP437). This should resolve the issue I observed of QWK-posted messages on FidoNet with the wrong CHRS header value (i.e. CP437, when the message body in fact contained UTF-8).

Some echolists (e.g. linuxnet.na) use '%' for comments (rather than ';') so just ignore any lines in imported echolists that begin with a non-alpha-numeric character.

*always* set the KILLSENT attribute flag on email/netmail received for relay via SMTP (regardless of the Internet/Fido netmail setting in SCFG).
The sender should already have a copy (e.g. in their outbox), if they want to keep it.

MSG_SENT is in the netattr field, not the attr field. Double-oops. Thanks Dan_C and Nelgin for bringing to my attention.

Reported by Deon George in https://sourceforge.net/p/syncterm/tickets/7/

- argv[0] does not always contain the full path to the executable, so that is
  not a good way to locate the exec directory.
- since we already need to know the ctrl dir and that is the normal place to
  put config files, that's where it shall be.
- jsdoor builds don't get a .ini file - I don't think they need it (?)

events aren't clicks (a click is button down then up, scroll events are
just button down).

Add support for scroll "buttons" (buttons 4 and 5).

When inkey() receives an ANSI escape sequence it doesn't need to consume, put the consumed characters into the keyboard buffer *ahead* of whatever else is in the keyboard buffer already. Uses new ungetkey() insert mode option.