Fixes https://github.com/behdad/harfbuzz/issues/63

We can't delete things before hb_ot_layout_position_finish().  So,
just zero the advance before it, and remove later.

Was broken by 82b521ae, as we have
positioning data by then and can't use the output buffer.  Ouch!

For example, fixes the following sequence with Arial XP:
628 25cc 651 25cc 64e 3a 20 628 651 34f 64e
628 25cc 64e 25cc 651 3a 20 628 64e 34f 651

Discovered as part of:
https://bugs.freedesktop.org/show_bug.cgi?id=85873

TODO: Documentation.

Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=71445

Based on patch from Martin Hosken, with review from Jonathan and I.

Based on https://bugzilla.mozilla.org/show_bug.cgi?id=1183209

https://bugs.freedesktop.org/show_bug.cgi?id=91228

Commit cdcdfe61 changed two `#pragma
message` to `#pragma error` in hb-unicode.cc, however MSVC uses #error,
just like the #else branch. `#pragma error` is an unknown pragma so
MSVC does not fail the build because of it, which I believe was the
intention of that commit.

If it's meant to be an #error, then the #ifdef for _MSC_VER can be
removed entirely.

Fixes test-shape, and:
https://code.google.com/p/chromium/issues/detail?id=497578

Separate the loops for the two cases of replacing with space
and deleting.  For deleting, use the out-buffer machinery.

Needed for upcoming cluster merge fix.

Based on test from https://code.google.com/p/chromium/issues/detail?id=497578

Currently fails.  Basically, if there's a default_ignorable at the
start of text, and font has no space glyph, we remove the default_ignorable,
and that makes the first char in text to correspond to no cluster.

Fix coming.

Until I figure out what to do about the API, and finalize
the tool.

Part of https://github.com/behdad/harfbuzz/issues/91

First try...

https://bugzilla.mozilla.org/show_bug.cgi?id=1167119

Fix "Since:" tags

Based on data from http://upstream-tracker.org/versions/harfbuzz.html
Resolves #103

Applying unary minus operator to unsigned int causes the following
warning on MSVS:

  warning C4146: unary minus operator applied to unsigned type, result still unsigned

Based on patch from Koji Ishi.

Fixes https://github.com/behdad/harfbuzz/pull/110

Using latest gobject-introspection, I don't seem to be having this
problem anymore:

  https://bugzilla.gnome.org/show_bug.cgi?id=707656

Removing that kludge makes language_t behave more like the way I expect it
in Python.

Also fixes:
https://github.com/behdad/harfbuzz/issues/91

Fixes user_data_t

Fix Travis CI config to pass again