feat: add MonthDay converter family with round-trip tests - #1325

Thanks @213f08 — the implementation matches the #1017 proposal (String-only, ISO --MM-dd with an MM-dd fallback, @DateTimeFormat support) and the registration mirrors YearStringConverter (all-converter, default-write by type, default-write by type + STRING).

Verification (worktree at ff051ba2, JDK 17, Windows)

Green:

• targeted: MonthDayStringConverterTest 16/16 + MonthDayConverterDataTest 15/15 = 31
• full: fesod-common 60 + fesod-sheet 1,820 = 1,880 tests, 0 failures
• spotless:check ok

Red:

• reverting only the DefaultConverterLoader registration (keeping the converter class) makes the registration test NPE and 12 of the 15 round-trip cases fail with Can not find 'Converter' support class MonthDay; restoring turns them green.

Non-blocking

1. The configured locale is ignored for @DateTimeFormat patterns. The sibling string converters pass globalConfiguration.getLocale() down to DateUtils (YearStringConverter, LocalDateStringConverter, LocalTimeStringConverter), while DateTimeFormatter.ofPattern(format) here always uses the JVM default locale. Probe (not committed): on a zh_CN JVM, with GlobalConfiguration.locale = Locale.US and @DateTimeFormat("MMMM d"), both writes produced the JVM-default Chinese month name and reading March 15 threw DateTimeParseException; reading 三月 15 succeeded only because the default locale happened to match. Numeric patterns such as MM/dd are unaffected. To align, either add DateUtils.parseMonthDay / format(MonthDay, ...) helpers (they also reuse the locale-keyed formatter cache) or pass the locale when building the formatter, plus one locale round-trip test.
2. Docs: after rebasing onto current main, please add a MonthDay row to website/docs/sheet/help/built-in-converters.md and the zh-CN counterpart (the page added in #1275 lists each converter family).
3. The branch is based on an older main (126dde9d) and I don't see CI results on GitHub yet. A rebase will surface the docs page and the new Windows CI matrix; worth pushing once so the checks run.

The rest looks good — leap day --02-29, invalid values (--02-30, month 13, full dates), blank/null pattern fallback, and xlsx/xls/csv round-trips are all covered. LGTM from my side, non-blocking items aside.