Update graphemes for Unicode 7

[stevengj]

This fixes #19, and also makes it much easier to implement grapheme iterators in Julia (JuliaLang/julia#9261) by adding a bool utf8proc_grapheme_break(int32_t c1, int32_t c2) function to check for a grapheme break between two codepoints. This allows us to iterate over graphemes in-place, without mapping to a separate string with 0xFF grapheme separators.

Unfortunately, I had to break backwards compatibility by changing the utf8proc_property_t struct to replace the extend:1 field with a boundclass:4 field, where the latter is now read from Unicode's GraphemeBreakProperty.txt file by the updated generator script. I took this opportunity to rearrange the struct to put the bitfields at the end, so that C will not insert alignment padding into the struct; as a consequence, the struct actually got smaller by several bytes.

Once this is merged, I will submit the corresponding patch to the utf8proc folks.

@jiahao, does it look okay to you? cc @StefanKarpinski

[StefanKarpinski]

👍

[stevengj]

Going to go ahead and merge, then submit upstream.