Xlocale for NetBSD

[Konrad Schroder <perseant%hhhh.org@localhost>]

I've made a fair start at porting FreeBSD's (port of Darwin's) xlocale 
to NetBSD.  This started out as a frustration at ${WORK}, but continued 
at home after we gave up at ${WORK} and switched OSs instead to get 
working LC_COLLATE support.

The benefits of adopting xlocale would be:

 * Per-thread locale settings
 * Working LC_COLLATE settings
 * Use localedef(1) instead of mklocale(1), which allows...
 * Take locale definitions directly from CLDR

but, at least at the moment, has these drawbacks:

 * No ISO2022 codemap

that is, no ja_JP.ISO2022* locales; iconv, which is still based on 
Citrus, continues to work properly so far as I know.

The lack of working ISO2022 really surprised me; the problem seems to be 
that the CLDR tools can't generate a working codemap for ISO2022.  I 
don't really know why that would be.

I have it building release and passing the few existing tests, except 
that it accepts previously forbidden sequences in Big5 encoding and 
chokes when you ask for ja_JP.ISO2022*. (Another possible issue is that 
the Citrus tests allow and test for 6-byte UTF-8 sequences, which are, 
strictly speaking, forbidden but not otherwise incompatible. I solved 
this problem by changing the code to pass the test, but it's easy to 
switch back if we want to go with the straight UTF-8 spec.)

I haven't applied the FreeBSD tests yet, and it needs new tests.  
(Actually, I haven't yet run a *complete* atf test run and compared 
results, so I may have missed some existing tests.  Easy enough to do.)

The current diffs are available at

  http://www.hhhh.org/perseant/xlocale-20170425.diff

There are some additional issues I'm aware of with this particular set 
of diffs but am intending to fix, to wit,

* No compatibility support; e.g. existing test binaries crash
* Missing locale(5) man page taken from Illumos; licensing issues?
* Make sure the man pages match the code
* Haven't tried the installer

If there's sufficient interest in having this in-tree, I'll commit it on 
a branch and continue development there.

Thoughts?

						Konrad Schroder
						perseant%hhhh.org@localhost