empty element test is wrong

[NSoiffer]

The resolution of #136 means that the test/code in presentation-markup/scripts/empty-underover.html is wrong. Those tests assume that having no children means the element takes 0 space. Given that these elements retain their layout based on #136 resolution, there are some min gaps and shifts that cause the elements to have some size (true for msub, munder, etc.)

[fred-wang]

Just for the record, we haven't decided anything precisely in #136: "Always render something to the user, at minimum mrow-like layout. Experiment more to see if we can do something that does not alter the visual rendering (Rob already did that for mfrac)."

I'm personally in favor of keeping mrow-like layout for now

[fred-wang]

This can be closed. https://wpt.fyi/results/mathml/presentation-markup/scripts/empty-underover.html pass in all browsers:

• Firefox 106 and Chrome 109 follows the spec and perform layout as an mrow with empty scripts.
• WebKit perform layout as an empty box.