Support nonlinear optimization step

[HannesSommer]

This requires:

• Output measurements CSVs with outliers replaced by empty lines / lines with timestamp only to feed to the NLOpt step
• Output the resulting extrinsic calibration (including temporal offset) as files (decide on format, should be readable by the user. Candidates: YAML directly or Boost property tree + XML / YAML / Info / Ini)
• Somehow communicate the type of sensor (relative or absolute pose measurements)

[NikolausDemmel]

How would this be different from what is already implemented in batch_estimator? Just the transfer of outliers?

[HannesSommer]

It is about improving the communication of the python scripts with the batch_estimator. The empty check boxes above indicate what is still missing in particular.