Settle distance moves faster with a minimum push near the target

[kevinyhe]

Why

The distance PID's output shrinks with the distance left. Near the target it falls to about what the drive needs to overcome friction, so the robot crawls. Then it exits on the "within 1.5 in for 250 ms" rule, short of the target. In the physics simulator, a 24 in drive on a 200 rpm base spent its last 3 in at 0.6–1.2 V, took 3.9 s and stopped 1.3 in short.

Change

• MotionConfig::settle_min_voltage (new, 1.5 V; 0 turns it off): stopped driveTo() and moveToPoint() push with at least this voltage, toward the target, while outside distance_exit.small_error (0.5 in).
  • moveToPoint(): the floor only covers the part of the distance left that lies along the heading, so a target off to the side isn't pushed into an orbit.
  • Inside 0.5 in, the PID alone finishes the move, so the floor can't make it hunt across the target.
  • Chained moves (exit false) are unchanged.
• boomerang() hysteresis: after it starts its final turn, it goes back to driving only when drift exceeds twice the position band. Before, it went back at the band edge, so a final turn that shifted the robot slightly could loop pivot, approach, turn until the time limit. That's what made stress/ramsete_low_grip regress with the floor alone (RAMSETE's settle ends with a boomerang).

Results

Full simulator matrix. No baseline scenario fails. The baseline is updated.

	Passing	Mean time, scenarios passing before and after	Mean miss
Two tracking wheels	99 → 101 / 118	2.79 → 2.41 s	1.07 → 0.60 in
Drive encoders only	82 → 84 / 118	2.49 → 2.13 s	1.45 → 0.81 in

24 in drives with tracking wheels:

Drive	Before	After
four_motor_200	3.91 s, 1.32 in	2.81 s, 0.56 in
six_motor_450	2.16 s, 1.26 in	1.38 s, 0.03 in
speed_base	1.28 s, 1.09 in	1.09 s, 0.20 in

Newly passing:

• four_motor_200/point_diagonal with tracking wheels, plain and with the fusion;
• six_motor_450/point_diagonal and speed_base/ramsete_s_curve with drive encoders.

Not fixed

four_motor_200/boomerang and speed_base/boomerang still time out, now at 43° and 102° off heading. The cause is how the boomerang approaches, not settling; it's #101. Steering to the final heading inside 6 in was tried and made every boomerang worse, so it's not in this PR.

Tests

• New in motion_safety_test: a drive that needs 1 V to start moving. Without the floor, driveTo(24 in) stalls at 21.48 in and runs out its 4 s. With it, the drive reaches 23.15 in in 2.84 s. 1216 checks.
• make test: 56/56 pass.