GEODE-10490: Clean up partial DS on retry and fix null pointer in test cleanup

1. LocatorStarterRule: Added cleanup of partial distributed system before retry.
   When startLocatorAndDS() fails with port binding error, it may have partially
   created a DS. This causes 'distributed system already exists' error on retry.
   Now properly stop locator and disconnect DS before retrying.

2. HeadlessGfshIntegrationTest: Added null checks in cleanup() method.
   If setup() fails (e.g., port binding failure), gfsh/cache/ds may be null,
   causing NullPointerException in @after cleanup. Added defensive null checks.

Author: JinwooHwang

geode-dunit/src/main/java/org/apache/geode/test/junit/rules/LocatorStarterRule.java

@@ -113,6 +113,18 @@ public void startLocator() {
 
       } catch (IOException e) {
         lastException = e;
+
+        // Clean up any partial distributed system that may have been created
+        try {
+          if (locator != null) {
+            locator.stop();
+            locator = null;
+          }
+          disconnectDSIfAny();
+        } catch (Exception cleanupException) {
+          // Ignore cleanup errors
+        }
+
         // Check if this is a port binding failure
         if (e.getMessage() != null && e.getMessage().contains("Address already in use")
             && attempt < maxRetries) {

geode-gfsh/src/integrationTest/java/org/apache/geode/management/internal/cli/HeadlessGfshIntegrationTest.java

@@ -91,9 +91,15 @@ private void legacyConnect(Properties properties) {
   @SuppressWarnings({"deprecation"})
   @After
   public void cleanup() {
-    gfsh.terminate();
-    cache.close();
-    ds.disconnect();
+    if (gfsh != null) {
+      gfsh.terminate();
+    }
+    if (cache != null) {
+      cache.close();
+    }
+    if (ds != null) {
+      ds.disconnect();
+    }
   }
 
   @Test