ECMA-376 §22.1.2 defines m:sty (style: p/b/i/bi) and m:scr (script: roman,
script, fraktur, double-struck, sans-serif, monospace) on math run
properties. Word emits m:sty="p" on function names like "lim", "max",
"sup" — without this handling those names rendered italic instead of
upright when they appear outside an m:func wrapper.

Precedence: m:sty > m:scr > m:nor. m:nor stays as the legacy normal
flag, equivalent to m:sty="p" when no explicit style or script is set.

- Tighten "handles missing m:e/m:lim gracefully" tests: assert arity-2
  <munder>/<mover> with an empty <mrow> on the missing side (was only
  checking the present side).
- Add multi-run base tests for both converters — existing tests only
  exercised multi-run m:lim, leaving the m:e base-wrapping path
  untested.
- Add nested-math tests (fraction inside m:lim) for both converters.
- Add m:func-wrapped tests (m:func > m:fName > m:limLow/m:limUpp) —
  the shape Word actually emits when the base is a function keyword.
- Add full coverage for m:sty (p/b/i/bi), m:scr (6 variants), and
  m:sty > m:scr precedence per ECMA-376 §22.1.2.

…/2378)

New fixture math-limit-tests.docx contains 8 Word-native limit shapes:
m:limLow in m:func, bare m:limUpp, m:limLow with fraction in m:lim,
bare m:limLow, m:limUpp in m:func, multi-char non-lim function bases
(max, sup), and m:limLow with a nested m:sSub in the limit expression.

Seven new it blocks assert:
- arity of <munder> (6) and <mover> (2)
- nested <mfrac> inside <munder> (case 3)
- nested <msub> inside <munder> (case 8)
- mathvariant=normal on function keywords via m:sty val=p
- bare identifiers keep italic default (no stray mathvariant)
- property elements (m:limLowPr/m:limUppPr/m:ctrlPr) don't leak to DOM

The same fixture is uploaded to the shared R2 corpus as
sd-2377-limlow-limupp-showcase-all-cases.docx so test:layout and
test:visual cover these shapes automatically.