Implements inertial_to_backend_frame_translation

[agalbachicar]

Pairs with maliput/maliput#401

[liangfok]

LGTM

[liangfok]

I am a little confused here. The docstring in https://github.com/ToyotaResearchInstitute/maliput/pull/401/files indicates that this method should only be overriden when "the Inertial and Backend Frames do not differ". Yet, the use of segment()->junction()->road_geometry()->inertial_to_backend_frame_translation() indicates that they may differ.

[liangfok]

Per discussion during meeting just now, if we provide dragway-specific support for an inertial-to-backend transform, it should have no run-time overhead. That is, the in-memory representation should be modified to account for the inertial frame. Thus, the only overhead should be at RoadNetwork initialization time.

[agalbachicar]

@liangfok note that dragway does not use geometry_base. All the linkage is managed by dragway::RoadGeometry constructor so we are only constrained by the API. I could revert this PR and just implement a stub RoadGeometry::inertial_to_backend_frame_translation() if that's more appropriate (same proposed solution in maliput_multilane --> maliput/maliput_multilane#46).

What do you think?

[agalbachicar]

We'll keep it as is and change it if required in the future.

[liangfok]

Okay.