fix(server): avoid loading huge task results for metadata queries

[contrueCT]

Purpose of the PR

Fix task metadata/recovery paths that can still load and decompress large task results, making GET /tasks, task restore, and DELETE /tasks/{id}?force=true unusable when historical tasks contain huge result payloads.

This is related to #3057 and #3059. The PR focuses on metadata-only access and cleanup paths, while large Gremlin result export/chunking can be handled separately.

Main Changes

• Add scheduler APIs that allow task reads with or without result payloads.
• Add metadata-only task deserialization that avoids reading ~task_result.
• Keep GET /tasks/{id} default behavior compatible, and add with_result=false.
• Make GET /tasks, task restore, and delete paths use metadata-only reads.
• Add lightweight task vertex deletion to avoid GraphTransaction.removeVertex() force-loading indexed task vertices.
• Avoid loading separated ~taskresult vertices when cleaning distributed task results.
• Return a result-stripped copy for in-memory tasks when withResult=false.

Verifying these changes

• Trivial rework / code cleanup without any test coverage. (No Need)
• Already covered by existing tests, such as (please modify tests here).
• Need tests and can be verified as follows:
  • mvn test -pl hugegraph-server/hugegraph-test -am -P core-test,rocksdb -Dtest=TaskCoreTest -DfailIfNoTests=false -Dcheckstyle.skip=true -Drat.skip=true
  • mvn compile -pl hugegraph-server/hugegraph-api -am -DskipTests=true -Dcheckstyle.skip=true -Drat.skip=true
  • mvn test-compile -pl hugegraph-server/hugegraph-test -am -P api-test,rocksdb -Dtest=TaskApiTest -DskipTests=true -Dcheckstyle.skip=true -Drat.skip=true

Does this PR potentially affect the following parts?

• Dependencies (add/update license info & regenerate_known_dependencies.sh)
• Modify configurations
• The public API
• Other affects (typed here)
• Nope

Documentation Status

• Doc - TODO
• Doc - Done
• Doc - No Need

[codecov]

Codecov Report

❌ Patch coverage is 59.84252% with 51 lines in your changes missing coverage. Please review.
✅ Project coverage is 29.90%. Comparing base (c3f56b5) to head (47fb2f9).

Files with missing lines	Patch %	Lines
.../apache/hugegraph/task/TaskAndResultScheduler.java	0.00%	26 Missing ⚠️
...g/apache/hugegraph/task/StandardTaskScheduler.java	70.37%	7 Missing and 1 partial ⚠️
.../main/java/org/apache/hugegraph/task/HugeTask.java	86.84%	2 Missing and 3 partials ⚠️
...pache/hugegraph/task/DistributedTaskScheduler.java	0.00%	4 Missing ⚠️
...ava/org/apache/hugegraph/task/TaskTransaction.java	76.47%	1 Missing and 3 partials ⚠️
.../org/apache/hugegraph/auth/HugeGraphAuthProxy.java	57.14%	3 Missing ⚠️
...ain/java/org/apache/hugegraph/api/job/TaskAPI.java	80.00%	1 Missing ⚠️

❗ There is a different number of reports uploaded between BASE (c3f56b5) and HEAD (47fb2f9). Click for more details.

HEAD has 2 uploads less than BASE

Flag	BASE (c3f56b5)	HEAD (47fb2f9)
	3	1

Additional details and impacted files

@@             Coverage Diff              @@
##             master    #3060      +/-   ##
============================================
- Coverage     36.10%   29.90%   -6.20%     
+ Complexity      338      264      -74     
============================================
  Files           803      804       +1     
  Lines         68291    68382      +91     
  Branches       8970     8982      +12     
============================================
- Hits          24656    20452    -4204     
- Misses        40990    45564    +4574     
+ Partials       2645     2366     -279     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[imbajin]

Blocking: yes. Summary: Found a scheduler API compatibility regression and a failing targeted test. Evidence: TaskCoreTest#testTaskWithoutResult failed locally; codecov/project is also failing.

---

🔗 Please also check the failed codecov/project status: https://app.codecov.io/gh/apache/hugegraph/pull/3060

[imbajin]

‼️ Preserve the existing tasks(...) result contract

The new TaskScheduler default methods keep the old behavior as tasks(ids) -> tasks(ids, true), but this override changes the task/result-backed scheduler to return metadata-only tasks by default. Existing Java callers of scheduler.tasks(...).next().result() can now silently get null while scheduler.task(id) still returns the result. Please keep the no-result path opt-in by removing these overrides or delegating them to withResult=true, and pass false only from the metadata-only call sites such as REST list, restore, delete, and scheduler scans.

[imbajin]

⚠️ Use a reloadable task callable in this regression test

This anonymous TaskCallable is persisted by class name, then the worker reloads it through TaskCallable.fromClass(). It has no public no-arg constructor, so the task fails to load as TaskCallable$1, and scheduler.waitUntilTaskCompleted(id, 10) times out. I reproduced it with mvn test -pl hugegraph-server/hugegraph-test -am -P unit-test -Dtest=TaskCoreTest#testTaskWithoutResult -DfailIfNoTests=false -ntp. Please use an existing reloadable callable or add a small static callable class for this test.