Here's how to use grapheme_extract() to loop across a UTF-8 string character by character.

<?php

$str = "سabcक’…";
// if the previous line didn't come through, the string contained:
//U+0633,U+0061,U+0062,U+0063,U+0915,U+2019,U+2026

$n = 0;

for (    $start = 0, $next = 0, $maxbytes = strlen($str), $c = '';
        $start < $maxbytes;
        $c = grapheme_extract($str, 1, GRAPHEME_EXTR_MAXCHARS , ($start = $next), $next)
    )
{
    if (empty($c))
        continue;
    echo "This utf8 character is " . strlen($c) . " bytes long and its first byte is " . ord($c[0]) . "\n";
    $n++;
}
echo "$n UTF-8 characters in a string of $maxbytes bytes!\n";
// Should print: 7 UTF8 characters in a string of 14 bytes!
?>

The other comments on this page were helpful for me.
However, consider using something better than empty($value) when checking the value returned by grapheme_extract since it could as well return something like "0" (which of course evaluates to false).

Looping through grapheme clusters:

<?php

// Example taken from Rust documentation: https://doc.rust-lang.org/book/ch08-02-strings.html#bytes-and-scalar-values-and-grapheme-clusters-oh-my
$str = "नमस्ते";
// Alternatively:
//$str = pack('C*', ...[224, 164, 168, 224, 164, 174, 224, 164, 184, 224, 165, 141, 224, 164, 164, 224, 165, 135]);
$next = 0;
$maxbytes = strlen($str);

var_dump($str);

while ($next < $maxbytes) {
    $char = grapheme_extract($str, 1, GRAPHEME_EXTR_COUNT, $next, $next);
    if (empty($char)) {
        continue;
    }
    echo "{$char} - This utf8 character is " . strlen($char) . ' bytes long', PHP_EOL;
}

//string(18) "नमस्ते"
//न - This utf8 character is 3 bytes long
//म - This utf8 character is 3 bytes long
//स् - This utf8 character is 6 bytes long
//ते - This utf8 character is 6 bytes long
?>