AVRO-4060: Use JDK to Hash Byte Array in UTF8

lang/java/avro/src/main/java/org/apache/avro/util/Utf8.java

@@ -174,8 +174,14 @@ public int hashCode() {
     if (h == 0) {
       byte[] bytes = this.bytes;
       int length = this.length;
-      for (int i = 0; i < length; i++) {
-        h = h * 31 + bytes[i];
+      // If the array is filled, use the underlying JDK hash functionality.
+      // Starting with JDK 21, the underlying implementation is vectorized.
+      if (bytes.length == length) {
+        h = Arrays.hashCode(bytes);
+      } else {
+        for (int i = 0; i < length; i++) {
+          h = h * 31 + bytes[i];
+        }
       }
       this.hash = h;
     }

lang/java/avro/src/test/java/org/apache/avro/util/TestUtf8.java

@@ -98,6 +98,26 @@ void hashCodeReused() {
     assertEquals(3198781, u.hashCode());
   }
 
+  /**
+   * There are two different code paths that hashcode() can call depending on the
+   * state of the internal buffer. If the buffer is full (string length eq. buffer
+   * length) then the JDK hashcode function can be used. This function can
+   * sometimes be vectorized JDK 21+ and therefore should be preferable. However,
+   * if the buffer is not full (string length le. buffer length), then the JDK
+   * does not support this and a scalar implementation is the only option as of
+   * today.
+   */
+  @Test
+  void hashCodeBasedOnCapacity() {
+    // string = 3; buffer = 3
+    Utf8 fullCapacity = new Utf8("abc");
+
+    // string = 3; buffer = 4
+    Utf8 partialCapacity = new Utf8("abcX").setByteLength(3);
+
+    assertEquals(fullCapacity.hashCode(), partialCapacity.hashCode());
+  }
+
   @Test
   void oversizeUtf8() {
     Utf8 u = new Utf8();