ICU-23360 Deduplicate subdivision suffix strings in ValiditySet

When ValiditySet constructs subdivisionData, it creates a separate String
object for each subdivision suffix via substring(). Suffixes like "01", "02",
"zzzz" repeat across many regions: of ~5,910 expanded entries, only 1,992
suffixes are unique, leaving 3,918 duplicate String objects (~195 KB).

Fix: call intern() on each suffix string. The JVM's intern pool ensures
that identical suffixes share a single String instance.

On Android, ValidityData is initialized in the zygote process, which is
forked to launch every app. Deduplicating these strings means the shared
pages backing them are less likely to be copy-on-written post-fork, so the
~195 KB saving is effectively per-process across every running app.

No parallel icu4c change needed. The C++ side doesn't have an equivalent
ValidIdentifiers/ValiditySet class; subdivision validity goes through a
different code path.

Author: jabagawee

icu4j/main/core/src/main/java/com/ibm/icu/impl/ValidIdentifiers.java

@@ -62,7 +62,7 @@ public ValiditySet(Set<String> plainData, boolean makeMap) {
                         pos2 = pos = s.charAt(0) < 'A' ? 3 : 2;
                     }
                     final String key = s.substring(0, pos);
-                    final String subdivision = s.substring(pos2);
+                    final String subdivision = s.substring(pos2).intern();
 
                     Set<String> oldSet = _subdivisionData.get(key);
                     if (oldSet == null) {