mark glyph width zeroing doesn't use unicode values when finding glyphInfo

[cmyr]

When converting a glyph from glyphs to ufo, we zero the widths of non-spacing marks. The logic here involves finding the category and subcategory if the glyph by looking it up in the bundled GlyphData; but this lookup only uses the user-assigned glyph name, which can be arbitrary.

Elsewhere when we fetch this data (such as when building public.openTypeCategories we use the name as well as the unicode values; we should be doing the same during conversion.

(I have a patch for this I'll push in a moment, just wanted an issue to reference)