OTF: account for robot rotation and use future-pose hub velocity

[phurley67]

Summary

• Use midpoint heading (trapezoidal approximation) for displacement prediction so the OTF loop correctly handles robot rotation during ball flight
• Compute shooter offset at the future heading since the shooter physically rotates with the robot
• Calculate hub angular velocity from the future pose/distance instead of the current pose, so the turret feedforward matches where we're actually aiming
• Update Swerve.getFuturePoseFromTime with the same midpoint-heading approach and correct twist omega

Note: This PR depends on #536 (fix/otf-spline-convergence).

Closes #537

Test plan

• Deploy and verify OTF shots land accurately when the robot is driving in a straight line (no rotation) -- behavior should be unchanged
• Test OTF shots while the robot is spinning (e.g., driver turning while moving). Shots should be noticeably more accurate than before
• Check that the "Future Pose" on Shuffleboard looks reasonable -- it should show the predicted position AND heading
• Verify convergence loop still converges within 15 iterations by logging iteration count
• Test at various distances (close, mid, far) while spinning to confirm the turret feedforward tracks smoothly

🤖 Generated with Claude Code