[OpenJDK 2D-Dev] RFR: 8190280 [macos] Font2DTest demo started failing for Arabic range from JDK 8 u162 b01 on Mac

[Philip Race]

Bug: https://bugs.openjdk.java.net/browse/JDK-8190280
Webrev: http://cr.openjdk.java.net/~prr/8190280/

This is a JDK8 version of the bug 
https://bugs.openjdk.java.net/browse/JDK-8172967
which affected harfbuzz in JDK 9

It is triggered by a couple of things but mainly that the Mac font code 
sometimes sets
a glyphid to be the negation of the unicode to tell a special fallback 
path that the
glyphid is not actually supported by the font and low-level rendering 
code should
supply a fall back.

The essence of the fix here is that in the ICU code we reject negative 
glyph ids
and treat them as missing.

The check for AIOBE in CompositeStrike is just an extra safe guard.

-phil