Skeleton-based formats in Chinese have duplicated 'month' glyph

This happens because we extend the width of the fields in `_adjust_skeleton` to match the originally-requested width.

- we look up a skeleton for `MMMMyyyy`; in case of chinese that's `yMMMM`
- the format string for this skeleton in chinese is 'y年M月'
- `y` and `M` expand to just the year/month without any symbols, so using that skeleton as-is is fine.
- however, we extend the width based on what we requested which results in `yyyy年MMMM月`
- but `MMMM` expands to `十一月` (for november) so now combined with the skeleton, the `月` at the end gets duplicated

I'm not sure if this is something we can fix on our side at all (without adding an ugly hack that adds a special case for the chinese locale).

If I understand https://unicode-org.github.io/cldr/ldml/tr35-dates.html#Matching_Skeletons correctly, this is actually the expected behavior and the fix would be for the unicode data to include more appropriate patterns for chinese?

> Once a skeleton match is found, the corresponding pattern is used, but with adjustments. Consider the following dateFormatItem: `d MMM y`
> If this is the best match for `yMMMMd`, pattern is automatically expanded to produce the pattern `"d MMMM y"` in response to the request.

Start by tracing the `_adjust_skeleton` logic described in the issue and reproduce the `MMMMyyyy` case for the Chinese locale. Compare the expanded pattern with the relevant CLDR skeleton-matching guidance, then determine whether the fix belongs in the adjustment logic or Unicode locale data. Done means the duplicated `月` glyph no longer appears without regressing skeleton expansion.

Written by the indexing model from the issue text.