Indic reordering of pre-base reordering characters is backwards

[dscorbett]

The Unicode Standard, Version 15.0 clarifies that “For combining characters placed to the left of a base character, the combining characters start from the base character and are arranged leftward.” HarfBuzz’s USE shaper already does this.

The Indic shaper needs updating. Here is ⟨किॎ⟩ <U+0915, U+093F, U+094E>:

The Myanmar shaper is a modification of the Indic shaper so it also needs updating. Here is ⟨ကေႄ⟩ <U+1000, U+1031, U+1084>:

The Khmer shaper does not allow multiple pre-base vowels so this issue does not apply. It probably uses a copy of the Indic code for reordering too, though, so it still ought to be fixed.

[behdad]

The Myanmar shaper is a modification of the Indic shaper so it also needs updating. Here is ⟨ကေႄ⟩ <U+1000, U+1031, U+1084>:

I thought for that one we tested and the order was intentional. @jfkthame @mhosken

[behdad]

The Indic shaper needs updating.

This would be hard to do with our main sort. We would need to do a run after the sort and reverse the left-matras.