Fix closed-track false goal stop, planner wp clamp, and short velocity crash

[cursor]

Bug and impact

1. Closed-track false GOAL_ARRIVED / immediate stop — With GoalArrivalMonitor + StopExecAtGoalTask, the rising-edge detector started as "not arrived". On closed race lines (goal_point == start_point), default Yas Marina start pose is ~2.1 m from the goal (inside the 3 m radius), so the first tick fired GOAL_ARRIVED and stopped execution before any driving.

2. Plan waypoint step-back / OOB — Plan UI step-back from wp 0 passed wp=-1 into LocalPlanningStrategy.reset. Numpy negative indexing jumped localization to track end (~5251 m on Yas). Oversized wp values could IndexError in the Tk UI.

3. Control crash on short/empty velocity profile — Stanley / PID / Pure Pursuit indexed velocity[current_wp] with no length guard. A plan whose ReferenceSpeed is shorter than the path crashed mid-drive once current_wp passed the short array.

Root cause

• Goal monitor treated "already inside radius at t0" as a rising edge.
• reset(wp) did not clamp; UI allowed negative wp.
• Controllers assumed len(velocity) == len(path).

Fix

• Seed GoalArrivalMonitor on the first tick (and after reset) without notifying; real arrivals still fire on leave → re-enter.
• Clamp LocalPlanningStrategy.reset(wp) to [0, n-1]; Plan UI step-back floors at 0.
• Clamp velocity index in Stanley / PID / Pure Pursuit; empty profile → zero accel.

Validation

• pytest for goal-monitor seeding, planner reset clamp, and short/empty velocity profiles — 20 related tests passed.
• Concrete Yas Marina numbers: start–goal distance 2.10 m; pre-fix step-back from wp0 → s≈track_end, post-fix stays at s≈0.