Update multi-vehicle behavior planning flow

[Tayhirro]

Summary

This PR improves the CAV behavior planning and local planning flow for multi-vehicle scenarios.

Main changes:

• Update route completion handling so scenario loops can manage finished CAVs without terminating the process.
• Improve behavior-agent route planning, car-following, and push-destination fallback logic.
• Refill the local planner waypoint buffer before trajectory generation.
• Reduce duplicated CAV control-loop code across scenario scripts by reusing ScenarioManager.

Validation

• git diff --check
• git diff --name-only -- '*.py' | xargs python3 -m py_compile

[xuhan417]

Hi Tayhirro,

Many thanks for contributing to the OpenCDA! Your work looks solid at first glance.

Could you please provide a brief technical document of the PR introducing the logic and improvement of this PR? It helps the team to better understand the issue and test/merge your work. A high-level template is provided here.

Best,
Xu