The multithreading doesn't have a correct comportment for pose measurement.
It should be investigated as soon as possible. For the moment, we disabled multithreading.

In external sensor API, a measurement can be added to a list. However, timestamp isn't checked which can lead to doubled measurement in list.
These doubled measurement can make the interpolation fail or decrease performances.

The tab key can be used to propose the different launch files for a specific package.

* Offset transform was inverted in computations
* Documentation is fixed
* Rename GpsFrameId to MapFrameId and change default value to map
* [ROS2] GPS reference frame is usually normalized (ECEF, ENU, etc.) while SLAM reference frame (i.e. odom) is local : e.g., first LiDAR sensor pose. It is more common to define the offset as the transform from the normalized reference to odom in the tree
* [ROS2] The map->odom transform is updated not to be static as odom is expected to move in time

* Frame could be linked with TF tree as well...

The calibration TF published was badly set (wrong frame ID)

The frame ID of the ext pose message is the frame in which the pose is represented, not the tracked one (e.g. "odom" not "ins") so we need to add a frame ID "ins" to get the calibration. User will need to set ins in his tf tree.