Order Chakma split vowels in accordance with NFC

[dscorbett]

U+1112E CHAKMA VOWEL SIGN O and U+1112F CHAKMA VOWEL SIGN AU decompose to a below-base vowel sign followed by an above-base vowel sign, the opposite of the usual USE order. Currently, HarfBuzz reverses the decomposition, which is not a complete solution: it causes a dotted to be inserted when the input buffer is in NFD. My fix recategorizes all the below-base Chakma vowels as above-base and vice versa, as in Tibetan.

[punchcutter]

This doesn't make any sense to me and breaks more than it fixes. What needs to be done is to update gen-use-table.py with the overrides listed at https://docs.microsoft.com/en-us/typography/script-development/use#overrides-to-indicpositionalcategory
I thought we had already done that, but I guess not yet. Also, like Behdad mentioned at #539 (comment) we should move the overrides to data files.

[dscorbett]

You are right: that broke U+11134 CHAKMA MAAYYAA followed by a below-base vowel. I have fixed that.

I know about the USE’s override list (that is what I was referring to when I mentioned Tibetan): it overrides all Chakma vowels to be below the base, which means that both U+11131 U+11127 and U+11131 U+11127 are valid, even though they look identical. It is preferable to avoid ambiguity by enforcing one canonical order, so I recommend not following the USE list exactly.