Standardize coordinate frames

[theochemel]

We currently consider two frames: odom and vehicle_ned. That won't be enough to handle everything we want to do this year.

We will now have four frames: world, odom, mission, and vehicle.

• vehicle is whatever we define to be the vehicle frame (likely positioned at IMU, x forward, y right)
• odom is the frame in which state estimation reports and motion planning runs.
• world is the frame in which persistent object locations (buckets!) are tracked.
• mission is the frame that we "retare" when convenient so we can work in a specific coordinate frame.

The frames will be related as follows.

• State estimation publishes the transform between odom and vehicle, which starts as identity.
• Other future localization that tells us our position relative to a fixed object publishes the transform between world and odom to correct for state estimation drift.
• We define the transform between world and mission as convenient, i.e. as the transform from world to vehicle to retare to our current transform.

[micahreich]