fix: 🐛 Run smoke tests for the simulation's frame count instead of 1000

[ShisatoYano]

Summary

In test mode (show_plot = False), GlobalXYVisualizer.draw() always ran 1000 frames. The simulations only need span_sec / interval_sec + 1 frames (81-551 across all 37 simulations), so every smoke test did 2-12x more work than the simulation itself.

This PR runs time_params.get_frame_num() frames, the same number the animation uses.

Closes #86

Changes

• src/components/visualization/global_xy_visualizer.py: replace range(1000) with range(self.time_params.get_frame_num())
• test/test_global_xy_visualizer.py (new): regression test that counts update() calls with a fake object and checks they match get_frame_num() (fails with 1000 == 51 before the fix)

Impact

• No simulation has more than 1000 frames, so no test becomes longer
• Smoke tests don't assert on the state after main(), so fewer frames don't change their results
• simulation_finished() still fires on the last frame, so the final full-range axis limits are still applied

Test

• pytest: 96 passed locally (38m36s on Linux)
• The slowest tests are now the path planners that encode search GIFs (rrt 571s, dijkstra 446s, ...). That is tracked separately in Path planning smoke tests write search GIFs and overwrite tracked files #87.
• Compare the CI runtime with the latest main run (Linux 52m / macOS 58m / Windows 1h49m)

🤖 Generated with Claude Code