Fixes http://github.com/behdad/harfbuzz/pull/36

Test with:

hb-view /Library/Fonts/Zapfino.ttf ZapfinoZapfino --features=-dlig[7:] --shaper=coretext

As exercised by hb-coretext .notdef code.

Based on errors seen here:
http://testresults.qt-project.org/ci/QtBase_dev_Integration/build_03837/wince70embedded-armv4i-msvc2008_Windows_7/log.txt.gz

Fully untested.

Apparently those functions documented as sometimes returning NULL
actually exercise that right in OS X 10.10 Yosemite.  The scratch
was too small for that.  I *think* I fixed it, but haven't tested
as I don't have Yosemite.

Fixes https://bugs.freedesktop.org/show_bug.cgi?id=82306

Apparently they are not (advertised as?) safe on BSD systems.
We ignore the case of static libraries.

Whitelisted on glibc, Android, and MSVC / mingw.

https://bugs.freedesktop.org/show_bug.cgi?id=82246

Was broken in 615d00ea.

Fixes https://github.com/behdad/harfbuzz/pull/48

Micro-test added.

Add buffer var allocation asserts to a few key places.

The reason we turned it on is because Kazuraki uses it.  But that's
not reason enough.  Until the OpenType spec gets its act together re
adding design-direction to lookups, this is better user experience.

Previously, we expected users to provide BOT/EOT flags when the
text *segment* was at paragraph boundaries.  This meant that for
clients that provide full paragraph to HarfBuzz (eg. Pango), they
had code like this:

  hb_buffer_set_flags (hb_buffer,
                       (item_offset == 0 ? HB_BUFFER_FLAG_BOT : 0) |
                       (item_offset + item_length == paragraph_length ?
                        HB_BUFFER_FLAG_EOT : 0));

  hb_buffer_add_utf8 (hb_buffer,
                      paragraph_text, paragraph_length,
                      item_offset, item_length);

After this change such clients can simply say:

  hb_buffer_set_flags (hb_buffer,
                       HB_BUFFER_FLAG_BOT | HB_BUFFER_FLAG_EOT);

  hb_buffer_add_utf8 (hb_buffer,
                      paragraph_text, paragraph_length,
                      item_offset, item_length);

Ie, HarfBuzz itself checks whether the segment is at the beginning/end
of the paragraph.  Clients that only pass item-at-a-time to HarfBuzz
continue not setting any flags whatsoever.

Another way to put it is: if there's pre-context text in the buffer,
HarfBuzz ignores the BOT flag.  If there's post-context, it ignores
EOT flag.

https://bugzilla.mozilla.org/show_bug.cgi?id=1045139#c26
https://github.com/behdad/harfbuzz/commit/b276e897d17519a2c28f79d024904ac2bdd86bcf#commitcomment-7243499

Restructure lookup array to accommodate.

https://bugzilla.mozilla.org/show_bug.cgi?id=1045139#c18

https://bugzilla.mozilla.org/show_bug.cgi?id=1045139#c18

https://bugzilla.mozilla.org/show_bug.cgi?id=1045139#c16

The table can now compile independently too.  If we cannot make it work
on MSVC, we can always generate the data and distribute it.

The code now compiles cleanly with:

gcc -c -xc -std=c99 -Werror -pedantic hb-ot-shape-complex-arabic-win1256.hh
g++ -c -xc -std=c++1x -Werror -pedantic hb-ot-shape-complex-arabic-win1256.hh

See:
https://github.com/behdad/harfbuzz/commit/a97f537cec209649302899975d76ca2b2661da7a#commitcomment-7218736

Bug 1045139 - The Arabic text with "MS Sans Serif" font is rendered bad
https://bugzilla.mozilla.org/show_bug.cgi?id=1045139

This is only enabled on Windows platforms, and requires support from
Uniscribe to work.  But for clients that do hook up to Uniscribe, this
fixes shaping of Windows-1256-encoded bitmap fonts like "MS Sans Serif".

The code and table together have just less than a 1kb footprint when
enabled.

UNTESTED.  I might even have broken regular Arabic fallback shaping.

See comments.

Micro-test added.

Seems to be what Uniscribe does.

At this point I think it's work checking our default...

Fixes Bug 76767 - Zeroing of advance of 2nd component of multiple
substitution with SBL Hebrew
https://bugs.freedesktop.org/show_bug.cgi?id=76767

Micro-test added.

Looks like Unsicribe responds to the 'mymr' tag by zeroing marks
GDEF_LATE instead of generic-shaper UNICODE_LATE.  Implement that.

Fixes
Bug 81775 - Incorrect Rendering with harfbuzz-ng myanmar unicode
https://bugs.freedesktop.org/show_bug.cgi?id=81775

Micro-test added based on Padauk.

Patch from Maks Naumov.
Fixes https://github.com/behdad/harfbuzz/pull/22

If user doesn't check hb_feature_from_string() return value, we
don't want them to end up see the partially-parsed feature.

With this, I believe we accept CSS feature strings completely.

We now allow things like "dlig on" / "dlig=on".

It's defined in hb-private.h already.

Towards accepting CSS font-feature-settings strings.