Description:
------------
ASCII optimisation of grapheme_* functions count CR + LF sequence as 2 distinct graphemes but should count for 1 as defined by Unicode standard.

It may conduct to some strange results depending if the string contains a code point > 0x7F or not. Eg:
grapheme_strlen("\r\na") != grapheme_strlen("\r\né")

A workaround for now could be to append an invisible or whitespace character to the string, like:

function my_grapheme_strlen($string) {
    return grapheme_strlen($string . "\xef\xbb\xbf") - 1; // append ZERO WIDTH SPACE (U+200B)
}

Test script:
---------------
var_dump(grapheme_strlen("\r\n"));
var_dump(grapheme_substr(implode("\r\n", ['abc', 'def', 'ghi']), 5));

Expected result:
----------------
int(1)
string(7) "ef
ghi"

Actual result:
--------------
int(2)
string(8) "def
ghi"