improvement: Don't use format since it seems to break at some chars

Related to scalameta/metals#7003

scalameta/tokenizers/shared/src/main/scala/scala/meta/internal/tokenizers/LegacyScanner.scala

@@ -232,9 +232,22 @@ class LegacyScanner(input: Input, dialect: Dialect) {
       }
     }
 
+    /**
+     * The algorithm for calculating unicode is as follows:
+     *   - `ch | 0x10000` gets us a new binary number: 1 at 17 position, then zeroes and then the
+     *     actual char number.
+     *   - `Integer.toHexString(ch \| 0x10000)` gets us the hexagonal unicode plus the one from 17
+     *     position
+     *   - `Integer.toHexString(ch | 0x10000).substring(1)` drops that added one
+     *
+     * This way we don't have to prefix with 0s since they will be included in the hex number
+     * string.
+     */
     @inline
-    def reportIllegalCharacter(): Unit = curr
-      .setInvalidToken(s"illegal character '\\u${"%04x".format(ch)}'")
+    def reportIllegalCharacter(): Unit = {
+      val output = "\\u" + Integer.toHexString(ch | 0x10000).substring(1);
+      curr.setInvalidToken("illegal character '" + output + "'")
+    }
 
     (ch: @switch) match {
       case ' ' =>