[TEST] use random ascii instead of random unicode

Otherwise fields might not have values after analysis and the docCount
and hence the score will be unpredictable.

core/src/test/java/org/elasticsearch/search/query/MultiMatchQueryIT.java

@@ -654,7 +654,7 @@ public class MultiMatchQueryIT extends ESIntegTestCase {
 
     public List<String> fillRandom(List<String> list, int times) {
         for (int i = 0; i < times; i++) {
-            list.add(randomRealisticUnicodeOfCodepointLengthBetween(1, 5));
+            list.add(randomAsciiOfLengthBetween(1, 5));
         }
         return list;
     }