Backport Density-Based Sampling to 3b CSM

[elvout]

Our 3b implementation systematically samples LaserScan messages, which theoretically causes the same problems as in the particle filter (#79).

• [x] Create a general "Observation" class in the models namespace to house raw scan readings (current particle_filter::Observation implementation) and corresponding point clouds (as Eigen matrices for efficient transformations).
• [x] Move density-based sampling to models as well. (as a method of the class maybe?)
• [x] Rework particle filter.
• [x] Compare reworked particle filter runtime distribution.
• [x] Rework CSM.

[elvout]

Once we have accuracy evaluation running, we can compare the accuracy of systematic sampling vs density based sampling.