Fix MathML spacing rules for mo elements. (mathjax/MathJax#2392)

[dpvc]

This PR modifies the MathML spacing rules for <mo> (and embellished <mo>) to correspond to Neil's explanation of the rules in mathjax/MathJax#2392. This means that space is only added when the (embellished) operator is inside an mrow (or inferred mrow) that has more than one element. This also makes the spacing between two successive <mo> elements be the maximum of the rspace of the first and the lspace of the second (as opposed to the sum of the two, as in the past).

This only affects when MathML spacing has been selected for the output jax, not the TeX spacing.

Finally, this also fixes the coreParent() function, which used to skip over <mstyle> elements, and so would incorrectly report the parent for <mstyle><mo>+</mo></mstyle> as the <mo> element rather than the <mstyle>. Similarly, the isTopEmbellished() function incorrectly reported the <mo> as the top embellished operator.

Resolve issue mathjax/MathJax#2392.

[zorkow]

Two things:

• This only applies to mrow elements? Not any other "implied" mrow?
• I have to admit I cannot really see the difference, but maybe I misunderstand what Neil's issue is.

Finally I think it is just wrong to ascribe meaning to a grouping element like mrow! But then again that is a major fault of MathML.

[zorkow]

f2f: lgtm.

[dpvc]

For the historical record:

This only applies to mrow elements? Not any other "implied" mrow?

The isKind('mrow') is actually an instanceof check under the hood, so it will be true for subclasses of mrow as well. The inferredMrow is a subclass of mrow, it will also make that test true. So the change applies to both explicit and implicit mrows.

I have to admit I cannot really see the difference, but maybe I misunderstand what Neil's issue is.

It only affects the output when mathmlSpacing: true is set in the output jax options, as it is about the MathML spacing rules, not the (default) TeX spacing rules.