[aat] Cluster index stays zero past `space` boundaries for three words of Hebrew with Lucida Grande

[drott]

We're failing fast/text/atsui-spacing-features.html in Chrome which tests justification and word spacing by inserting space and cluster boundaries, etc. - link to test failure here.

CoreText reports a cluster index increment from 0 to 2 after yodhebrew, shevanarrowhebrew, and keeps incrementing the cluster index, next after hiriqhebrew, etc.

$ ../test/shaping/hb-unicode-encode U+05D9,U+05B0,U+05D4,U+05B4,U+05D9,U+002C,U+0020,U+05D0,U+05B8,U+05D7,U+05B4,U+05D9,U+002C,U+0020,U+05DC,U+05B0,U+05DA,U+05B8,U+0020 | ./hb-shape --shaper=coretext --font-funcs=ot /System/Library/Fonts/LucidaGrande.ttc
[space=18+648|finalkafqamatshebrew=16+1081|shevahebrew=14+0|lamedhebrew=14+1092|space=13+648|comma=12+648|yodhebrew=11+650|hiriqhebrew=9+0|hethebrew=9+1458|qamatshebrew=7+0|alefhebrew=7+1321|space=6+648|comma=5+648|yodhebrew=4+650|hiriqhebrew=2+0|hehebrew=2+1458|shevanarrowhebrew=0+0|yodhebrew=0+650]

However, when shaping with HB's own AAT shaper, the cluster index is not correctly reported and stays at zero up until the end of the last word of the test string, then jumps to the correct 18 for the trailing space.

$ ../test/shaping/hb-unicode-encode U+05D9,U+05B0,U+05D4,U+05B4,U+05D9,U+002C,U+0020,U+05D0,U+05B8,U+05D7,U+05B4,U+05D9,U+002C,U+0020,U+05DC,U+05B0,U+05DA,U+05B8,U+0020 | ./hb-shape --shaper=ot --font-funcs=ot /System/Library/Fonts/LucidaGrande.ttc
[space=18+648|finalkafqamatshebrew=0+1081|shevahebrew=0@-102,-22+0|lamedhebrew=0+1092|space=0+648|comma=0+648|yodhebrew=0+650|hiriqhebrew=0@46,0+0|hethebrew=0+1458|qamatshebrew=0@12,-17+0|alefhebrew=0+1321|space=0+648|comma=0+648|yodhebrew=0+650|hiriqhebrew=0@46,0+0|hehebrew=0+1458|shevanarrowhebrew=0@-180,0+0|yodhebrew=0+650]