Gym mecanum kinematics matching to Webots roller-hinge proto

Mecanum proto rewrite in b3cf990 made the wheels truly omnidirectional
in Webots, but with asymmetric slip: forward command produces ~89% of
textbook speed while strafe produces only ~38% plus a consistent
~28% backward bleed-through. v1 BC/RL trained on perfect mecanum
gym kinematics could not herd the new dynamics. To unblock that:

* `mecanum_kinematics_step` gains two parameters that scale the
  realised motion to match a deployed-platform calibration:
    - strafe_efficiency  ∈ (0, 1]  default 1.0
    - strafe_to_forward_bleed     default 0.0
  Forward motion is untouched (textbook X-pattern continues to apply
  to vx_body); only the lateral channel is scaled and bleed is added.
* `RobotConfig` exposes both as drive-config fields with the same
  pass-through defaults so existing diff-drive code and existing
  mecanum training pipelines see no behaviour change.
* `HERDING_MEC_WEBOTS` preset bakes in the values measured against the
  current Webots mecanum proto (strafe_efficiency=0.4,
  strafe_to_forward_bleed=-0.28). Training mecanum BC/RL with this
  preset produces policies that compensate for the imperfect
  physical mecanum at deploy.
* `HerdingEnv` plumbs `RobotConfig.strafe_*` through to
  `mecanum_kinematics_step` so the preset takes effect.
* tools/gen_mecanum_wheels.py is added so the proto's 32 roller
  hinges can be regenerated by editing a single set of constants
  rather than hand-editing 1500+ lines of VRML.

Tests:
* 4 new mecanum_kinematics_step tests (default pass-through, strafe
  scaling, backward bleed, forward unaffected by strafe params).
* 3 new RobotConfig tests (defaults, validation, preset shape).
* Sanity check: gym strafe with HERDING_MEC_WEBOTS over 100 steps
  reproduces the Webots calibration to 2 decimal places.

126 unit tests pass (was 120).

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

herding/config.py

@@ -252,11 +252,32 @@ class RobotConfig:
     sees at deployment.
     """
 
+    strafe_efficiency: float = 1.0
+    """Mecanum strafe magnitude as a fraction of textbook X-pattern.
+
+    ``1.0`` (default) = perfect mecanum kinematics. ``0.4`` matches the
+    Webots roller-hinge mecanum proto calibration (62% slip on strafe,
+    11% on forward). Used by ``mecanum_kinematics_step`` only — has no
+    effect on differential drive.
+    """
+
+    strafe_to_forward_bleed: float = 0.0
+    """Fraction of ideal strafe magnitude that bleeds into body-frame x.
+
+    ``0.0`` (default) = no bleed. ``-0.28`` matches the Webots proto's
+    consistent backward push under strafe commands. Used by
+    ``mecanum_kinematics_step`` only.
+    """
+
     def __post_init__(self) -> None:
         if not (0.0 <= self.action_smooth < 1.0):
             raise ValueError(
                 f"action_smooth must be in [0, 1), got {self.action_smooth}"
             )
+        if not (0.0 < self.strafe_efficiency <= 1.0):
+            raise ValueError(
+                f"strafe_efficiency must be in (0, 1], got {self.strafe_efficiency}"
+            )
 
     @property
     def max_linear(self) -> float:
@@ -360,6 +381,39 @@ HERDING_WEBOTS = HerdingConfig(
     ),
     robot=RobotConfig(action_smooth=0.55),
 )
+
+HERDING_MEC_WEBOTS = HerdingConfig(
+    lidar=LIDAR_WEBOTS,
+    detection=DetectionConfig(wall_reject=0.5, static_reject=1.2),
+    tracker=TrackerConfig(
+        forget_steps=300,
+        max_new_tracks_per_step=1,
+        pen_latch_depth=2.0,
+        predict_steps=180,
+        consensus_k=3,
+        consensus_radius_m=0.3,
+        consensus_max_age=20,
+    ),
+    robot=RobotConfig(
+        action_smooth=0.55,
+        strafe_efficiency=0.4,
+        strafe_to_forward_bleed=-0.28,
+    ),
+)
+"""Webots-mecanum-matched training preset.
+
+Same as HERDING_WEBOTS but with the gym mecanum kinematics scaled to
+match the Webots roller-hinge mecanum proto:
+* ``strafe_efficiency=0.4`` — strafing produces ~40% of textbook
+  X-pattern lateral velocity in Webots; this matches the bias.
+* ``strafe_to_forward_bleed=-0.28`` — strafe commands bleed ~28% of
+  their magnitude into backward body motion in Webots.
+
+Use this preset when training BC/RL for the mecanum drive so the
+policy learns to compensate for the imperfect physical mecanum.
+Differential drive ignores both parameters and behaves identically
+to HERDING_WEBOTS.
+"""
 """Webots-matched training preset.
 
 Changes vs HERDING_DEFAULT: