[ISSUE #7701] Fix scaling monitor scheduler cleanup on context close.

[ymy4171]

Fixes #7701.

Closing the Spring application context previously left the scaling monitor's executor running because TaskSchedulerManager creates its scheduler internally and its existing shutdown() method was not registered as a destruction callback.

Add @PreDestroy to the existing cleanup method so context shutdown cancels registered tasks and shuts down the scheduler. Add real Spring-context regression tests covering both an active monitor task and a monitor task cancelled before context close; both assert executor shutdown and thread termination.

Validation:

• Both new regression cases failed without the callback and passed with it.
• On the rebased branch: Maven Wrapper 3.6.3 / JDK 21, ./mvnw -pl shenyu-admin verify (Windows wrapper, using local Maven mirror settings): 1998 tests, 0 failures, 0 errors, 1 skipped.
• Checkstyle, Apache RAT, and git diff --check passed.
• The full reactor was not rerun after rebasing. An earlier full-reactor baseline at b53408861 failed in the unrelated ApolloDataServiceTest.testProxySelectorDataDeleted test.

Checklist:

• I have read the contribution guidelines.
• I have added regression tests for the change.
• Full-repository ./mvnw clean install -Dmaven.javadoc.skip=true passed; module validation and the earlier baseline limitation are described above.

[Aias00]

Approved.

Adding @PreDestroy to TaskSchedulerManager.shutdown() correctly releases the monitor scheduler and cancels in-flight tasks when the application context closes, fixing the cleanup gap described in #7701. The new TaskSchedulerManagerTest drives the full Spring lifecycle (AnnotationConfigApplicationContext close) and asserts the executor terminates.

Minor note: CI checks have not reported on this branch yet; please confirm the workflow passes before merging.