Description:
------------
An incorrect behavior of the function was detected when working with text in the UTF-8 format. 
Instead of changing characters, characters disappear.
The string must be interrupted by a carry, and in fact, it is replaced with characters from another encoding.

Test script:
---------------
<?php

$a = "Это тестовая строка";

var_dump( preg_replace( '/(те)сто(вая)/u', "\1\n\2", $a ) );
var_dump( preg_replace( '/(те)сто(вая)/u', "\1" . "\n" . "\2", $a ) );
var_dump( preg_replace( '/(те)сто(вая)/u', '\1' . "\n" . '\2', $a ) );
?>

Expected result:
----------------
string(23) "Это те
вая строка"
string(23) "Это те
вая строка"
string(31) "Это те
вая строка"

Actual result:
--------------
string(23) "Это 

 строка"
string(23) "Это 

 строка"
string(31) "Это те
вая строка"