Update GH CI (#1658)

As test was naively using `new File("")` assuming its method `file.list()` will return `null` on cleanup in `teardown()` method.

But, it turns out it was an old Java bug fixed in
Java 25, and once GH matrix got Java 25 introduced, it broke the build.

Refs:
* https://bugs.openjdk.org/browse/JDK-8024695
* https://inside.java/2025/11/03/quality-heads-up/

Thanks to @sormuras

Author: cstamas

.github/workflows/maven-verify.yml

@@ -28,7 +28,6 @@ jobs:
     with:
       ff-run: false
       ff-site-run: false
-      jdk-matrix: '[ "21" ]'
-      maven-matrix: '[ "3.9.11", "4.0.0-rc-4" ]'
+      maven4-enabled: true
       verify-fail-fast: false
-
+      jdk-matrix: '[ "21", "25" ]'

maven-resolver-impl/src/test/java/org/eclipse/aether/internal/impl/DefaultArtifactResolverTest.java

@@ -772,7 +772,7 @@ void testFindInLocalRepositoryWhenVersionWasFoundInLocalRepository() throws Arti
         session.setLocalRepositoryManager(new LocalRepositoryManager() {
 
             public LocalRepository getRepository() {
-                return new LocalRepository(new File(""));
+                return new LocalRepository(lrm.getRepository().getBasePath());
             }
 
             public String getPathForRemoteMetadata(Metadata metadata, RemoteRepository repository, String context) {
@@ -841,7 +841,7 @@ void testFindInLocalRepositoryWhenVersionRangeWasResolvedFromLocalRepository() t
         session.setLocalRepositoryManager(new LocalRepositoryManager() {
 
             public LocalRepository getRepository() {
-                return new LocalRepository(new File(""));
+                return new LocalRepository(lrm.getRepository().getBasePath());
             }
 
             public String getPathForRemoteMetadata(Metadata metadata, RemoteRepository repository, String context) {