Interpolate camera and skater presentation between fixed ticks

docs/RENDER_INTERPOLATION.md

@@ -0,0 +1,24 @@
+# Render interpolation
+
+The Bevy presentation layer retains two completed simulation snapshots and
+interpolates them with `Time<Fixed>::overstep_fraction()` in `Update`. Capture
+runs after the physics system on every fixed tick, including catch-up ticks.
+Frames without a new simulation tick still advance the interpolation fraction.
+
+Snapshots contain the animation root, final physical skin hierarchy (including
+the board/trucks/wheels), and camera transform/FOV. Bone transforms are converted
+to skin-local transforms at each endpoint before translation/scale lerp and
+quaternion slerp. The root and camera use that same fraction. These rendered
+transforms are not read back by the native simulation or camera controllers.
+
+Initialization, camera discontinuities, the existing camera-reset flag and
+simulation-period changes collapse the history to the newest snapshot. There
+is no boxcar camera filter, additional smoothing window, or extrapolation.
+Interpolation introduces one fixed interval of presentation delay (normally
+about 16.7 ms). This is a renderer feature, not a claim about retail behavior.
+
+Reference: Bevy's official fixed-timestep presentation example:
+https://bevy.org/examples-webgpu/movement/physics-in-fixed-timestep/
+
+Validation: `Build.ps1` succeeded; no automated or gameplay tests were run,
+as requested. Gameplay smoothness remains for the user to assess.