Re: Checking pgwin32_is_junction() errors

Here's a better idea, now that I'm emboldened by having working CI for
Windows frankenbuilds, and since I broke some stuff in this area on
MSYS[1], which caused me to look more closely at this area.

Why don't we just nuke pgwin32_is_junction() from orbit, and teach
Windows how to lstat()? We're already defining our own replacement
stat() used in both MSVC and MSYS builds along with our own junction
point-based symlink() and readlink() functions, and lstat() was
already suggested in a comment in win32stat.c.

There's one curious change in the draft patch attached: you can't
unlink() a junction point, you have to rmdir() it. Previously, things
that traverse directories without ever calling pgwin32_is_junction()
would see junction points as S_ISDIR() and call rmdir(), which was OK,
but now they see S_ISLNK() and call unlink(). So I taught unlink() to
try both things. Which is kinda weird, and not beautiful, especially
when combined with the existing looping weirdness.

0001 is a copy of v2 of Melih Mutlu's CI patch[2] to show cfbot how to
test this on MSYS (alongside the normal MSVC result), but that's not
part of this submission.

[1] https://www.postgresql.org/message-id/flat/b9ddf605-6b36-f90d-7c30-7b3e95c46276%40dunslane.net
[2] https://www.postgresql.org/message-id/flat/CAGPVpCSKS9E0An4%3De7ZDnme%2By%3DWOcQFJYJegKO8kE9%3Dgh8NJKQ%40mail.gmail.com