Add workflow with thinking budget

[pan-x-c]

Description

Add a ThinkingBudgetWorkflow that:

• Reads thinking_token_budget from workflow_args.
• Uses raw_task["messages"] directly as OpenAI API messages.
• Passes the budget to vLLM through extra_body.
• Detects the forced </think> tokens returned by vLLM and masks them out in Experience.action_mask.
• Defaults reasoning_end_str to </think>, while allowing explicit overrides.

Tests cover budget forwarding and action-mask correctness.

Checklist

Please check the following items before code is ready to be reviewed.

• Code has passed all tests
• Docstrings have been added/updated in Google Style
• Documentation has been updated
• Code is ready for review

[pan-x-c]

/unittest-module-common

[github-actions]

unittest: Run #1827

Tests 📝	Passed ✅	Failed ❌	Skipped ⏭️	Pending ⏳	Other ❓	Flaky 🍂	Duration ⏱️
88	85	2	1	0	0	0	39m 48s

❌ Some tests failed!

Name	Failure Message
❌ tests/common/vllm_test.py::TestMessageProcess::test_truncation_status	The test failed in the call phase
❌ tests/common/vllm_test.py::TestQwen35APIServerReasoning::test_thinking_token_budget_action_mask	The test failed in the call phase

Github Test Reporter by CTRF 💚

[pan-x-c]

/unittest-module-common

[github-actions]

unittest: Run #1828

Tests 📝	Passed ✅	Failed ❌	Skipped ⏭️	Pending ⏳	Other ❓	Flaky 🍂	Duration ⏱️
88	85	2	1	0	0	0	37m 19s

❌ Some tests failed!

Name	Failure Message
❌ tests/common/vllm_test.py::TestQwen35APIServerReasoning::test_thinking_token_budget_action_mask	The test failed in the call phase
❌ tests/common/vllm_test.py::TestConcurrentSyncWeights::test_chat_during_weight_sync	The test failed in the call phase due to an assertion error

Github Test Reporter by CTRF 💚

[pan-x-c]

/unittest-module-common

[github-actions]

unittest: Run #1829

Tests 📝	Passed ✅	Failed ❌	Skipped ⏭️	Pending ⏳	Other ❓	Flaky 🍂	Duration ⏱️
88	87	0	1	0	0	0	39m 32s

🎉 All tests passed!

Github Test Reporter by CTRF 💚