Ignore variation selectors in operator category determination

[Jules-Bertholet]

Ensures that variation sequences in https://www.unicode.org/Public/latest/ucd/StandardizedVariants.txt are classified properly. For example, the sequence [U+2295 CIRCLED PLUS, U+FE00] (⊕︀) should have the same lspace and rspace than plain U+2295 CIRCLED PLUS (⊕).

This does not account for the case where a variation selector is followed by U+0338 COMBINING LONG SOLIDUS OVERLAY or U+20D2 COMBINING LONG VERTICAL LINE OVERLAY (resulting in a string of length 3). Further changes would be necessary to handle that edge case.

[w3cbot]

bert-github marked as non substantive for IPR from ash-nazg.

[Jules-Bertholet]

Perhaps what the spec should really say is that any Unicode combining character sequence should be equivalent to its base character.

[davidcarlisle]

@Jules-Bertholet Thanks for this, The PR has merge conflicts but the intention is clear (it might be better to first have an issue so we could discuss the exact version of the enhancement and then make a new PR rebased against the current head)

That said, we discussed this at today's MathML-core meeting, there was general support for the idea, but we would need to check implementation details to decide if this can be fitted in to this cycle or would need to be mathml-core level 2.