This Coverity reported issue was previously resolved (incorrectly) with
commit d02fc1a2 which also introduced a bug that ate all the unexpected
cursor position report characters (causing issue #304).

The correct fix was to compare the response length against the buffer size
minus one, to leave room for the NUL terminator.

While fixing this and issue #304, I noticed that this function was using the
response length ('rsp') as both the state machine state and string length,
which didn't work right if a nearly-matching report/response was received:
the x/y values would have been stripped from the response before stuffing
in the keyboard input buffer. So a bit of a refactor here using a proper
state machine variable.

Unexpected characters received (maybe typed by the user) were supposed to
be stuffed in the keyboard input buffer using ungetstr(), but the input string
was cleared rather than NUL-terminated before logging the expected character
and calling ungetstr(), so nothing (a blank string) was logged and nothing was
stuffed in the keyboard input-buffer in this case. Just an off-by-one bug.

This should fix issue #304. Thanks to Keyop and Nelgin for their persistence
in reporting and testing (in #synchronet at irc.synchro.net).

I didn't look closely enough at which write() call I was instrumenting.
This is now more useful, though I think we already found the root-cause
based on different log output ("Unexpected ansi_getxy response").

This just resulted in a lot of unhelpful noise in the log output

When reading and writing characters to chat files, log each with a debug-level
log message. If a read or a write fails, log with an error-level log message.

This will hopefully help root-cause and resolve issue #304.

... for proper error reporting, at least in the terminal server.

Also define/use a macro for the user base filename (now user.tab, not user.dat).

instead of USER_RECORD_LINE_LEN

and then unlocking again. So... make readuserdat() optionally leave the
record locked, reducing a lot of unnecessary lock/unlocking in here.

Move the user field index definitions to userfields.h

Also resolved some 32 vs 64-bit 'long' issues/ambiguities that have long-remained. :-)

This commit also removes logon.lst file support.

There's a TODO block remaining in js_user.c for setting portions of a user's birthdate (e.g. just the year or month or day).