Fix teleop re-engagement by resetting robot pose anchor on move start

[MNikoliCC]

Summary

This PR fixes move re-engagement behavior in lerobot_teleoperator_teleop.

The teleoperator now treats each new move press as a fresh teleop session and computes pose_from_initial relative to the pose at the start of that session. This prevents stale motion from being reused when move is released and pressed again.

What changed

In lerobot_teleoperator_teleop/teleop.py:

• track the pose at the start of the current move session
• reset that session reference when move is released
• compute pose_from_initial as a delta from the current session start pose

Why

Before this change, re-pressing move after releasing it could reuse previously accumulated motion instead of starting from zero for the new hold session.

With this change:

• press and hold move: motion is tracked normally
• release move: session state is cleared
• press move again: delta starts from zero for the new session

Scope

This PR only updates lerobot_teleoperator_teleop.

Robot-specific/custom interfaces that consume pose_from_initial may still need matching adjustments on their side depending on how they anchor and apply the pose delta.
Example: SpesRobotics/lerobot-robot-xarm#1 (comment)