At any position, if state is not zero, mark unsafe-to-break before,
unless we can reason it safe.

At any position, if there's an action entry for end-of-text, mark
unsafe to break.

Also changes buffer diff impl to allow for flag differences as long
as the buffer glyph flags are superset of reference glyph flags.

With this, all MORX tests pass.

See comment.

With this, MORX-20 passes if I turn --verify off.  Our unsafe-to-break
logic is currently broken in presence of end-of-text actions.  That's,
ugh, extra work to fix.  Let me try...

Fixes MORX-21.

Fixes MORX-18, MORX-19, and MORX-22.

I hope... Makes "ffi" work with Zapfino. I'm not sure if doing it correctly though...

We form the Zapfino ligature now. Yay! No further testing done.

I hate it that with my compiler at least, it doesn't catch totally uncompilable
statements in templates if they are not used...

Not sure why only one of the bots sees this as visible...

Maybe I should disable C++11 detection such that my compiler also catches
these.  Doesn't look like we can switch to it for a while anyway...

It's dead, Jim!

Apparently CoreText does allow these.  To be done: detect infinite
loops.

Fixes MORX-12 test.

Don't know why this is not caught by my compiler.  Looks like clang catches it
while gcc doesn't.

In file included from ../src/hb-aat-layout-morx-table.hh:31:
../src/hb-aat-layout-common-private.hh:523:30: error: no member named 'static_size' in 'AAT::ContextualSubtable<AAT::MorxTypes>::EntryData'
  DEFINE_SIZE_STATIC (4 + T::static_size);
                            ~~~^

Fixes MORX-11 test.

Tested with Zapfino and text "2nd". Sascha will create tests later.

Passes MORX-10 test now.

Neato.

To be improved later.

Such a headache that Windows defines UINT8, ...; Just prefix it.