Add the ability to reset map & keep map -> odom transform.

[Krzo99]

I think it would be usefull to be able to reset the whole map, but keep map -> odom transform, so new map would be drawn in Rviz2 where the last one ended. This could also be done the same way it's done with odom. A service, you can call to set map -> odom transform (reset_odom_to_pose).

Let me know what you think. Thanks.

[matlabbe]

Is it just for cosmetic purpose or for a multi-session workflow? The idea behind resetting the map is to restart from the current odometry pose. Unlike the odometry pose that we could arbitrary reset to any pose, the map->odom transform represents how much the graph has been corrected after loop closures, a map->odom with Identify tells us that there are no loop closures detected or if it stays close to Identity after loop closures, it would mean very low odometry drift. Faking a map->odom transform would mean that we add a loop closure correction that doesn't exist. The better way would be to have a node listening to map->odom, save the transform, reset the map so that map->odom will come back to Identity and publish back the saved transform under a new frame called world->map. If rviz's global frame is world, then if we reset the map from that intermediate node, the new map would appear restarting where we reset the map.

[Krzo99]

For multi-purpose mapping. The idea is, to reser the map, when mapping new areas, to speed up & lower the merged map data. But I like the idea of having a world->map transform, that you adjust for each robot, when you reset the map.

[matlabbe]

I think world->map approach would make more sense. By the way, you can use /rtabmap/load_database service to create a new database online for convenience if you were not using it.