not consuming full EGC for ZWJ

[dankamongmen]

I've added a new unit test loading

🤽🏼‍♀️ Woman Playing Water Polo: Medium-Light Skin Tone

into an nccell. We're only loading 11 bytes, but we ought be loading 17:

U+1F93D WATER POLO
UTF-8: f0 9f a4 bd UTF-16BE: d83edd3d Decimal: 🤽 Octal: \0374475
🤽
Category: So (Symbol, Other); East Asian width: W (wide)
Unicode block: 1F900..1F9FF; Supplemental Symbols and Pictographs
Bidi: ON (Other Neutrals)

U+1F3FC EMOJI MODIFIER FITZPATRICK TYPE-3
UTF-8: f0 9f 8f bc UTF-16BE: d83cdffc Decimal: 🏼 Octal: \0371774

Category: Sk (Symbol, Modifier); East Asian width: W (wide)
Unicode block: 1F300..1F5FF; Miscellaneous Symbols and Pictographs
Bidi: ON (Other Neutrals)

U+200D ZERO WIDTH JOINER
UTF-8: e2 80 8d UTF-16BE: 200d Decimal: ‍ Octal: \020015

Category: Cf (Other, Format); East Asian width: N (neutral)
Unicode block: 2000..206F; General Punctuation
Bidi: BN (Boundary Neutral)

U+2640 FEMALE SIGN
UTF-8: e2 99 80 UTF-16BE: 2640 Decimal: ♀ Octal: \023100
♀
Category: So (Symbol, Other); East Asian width: A (ambiguous)
Unicode block: 2600..26FF; Miscellaneous Symbols
Bidi: ON (Other Neutrals)

U+FE0F VARIATION SELECTOR-16
UTF-8: ef b8 8f UTF-16BE: fe0f Decimal: ️ Octal: \0177017

Category: Mn (Mark, Non-Spacing); East Asian width: A (ambiguous)
Unicode block: FE00..FE0F; Variation Selectors
Bidi: NSM (Non-Spacing Mark)

[dankamongmen]

uc_is_grapheme_break() definitely returns true for U+2640 FEMALE SIGN. i think we're misinterpreting the meaning of a grapheme break?

[dankamongmen]

it doesn't appear to be a context-sensitive thing, either, since u8_grapheme_next() also returns 11 bytes here

[dankamongmen]

if we check for \u200d and force a join following it, we get the expected 17 bytes. i'm not sure we want this...not all sequences are valid, and what if there are two such characters. if nothing else, this does seem to improve the appearance of mojibake...