The regexps are meant to extract a match immediately following the cursor - the
anchor should have been there from the beginning...
Also empty sub/superscript doesn't make sense - nested sub/superscript does
make sense but yagni.
Until now we expected the .org file to print back to itself - we can't do that
when the input is not pretty printed already - with the introduction of blocks
with unindented content that will be the case.