We should now avoid dumping the same data several time (in sensors LOS,
datation and rate), and we use linear interpolation on replay between
the dumped LOS, datation and rates.

The replay does not work correctly for now (the unit test fails), so
this needs more work.

The previous name was confusing as another mechanism for debugging also
use the term 'dump' for something not related to interpolator reuse.

Fixes issue #189

The new test for the ancestor method fails, but the failure is probably
related to issue#189 (in fact the issue was identified while testing
this). So we will first solve issue#189 in another branch and merge it
once solved, before continuing working on this branch.

This is an attempt at solving issue #188.

This algorithm is similar in spirit to the IGNORE_DEM_USE_ELLIPSOID, but
with a user-specified elevation instead of always using 0.0 elevation.

Implements feature #187.

Thanks to Espen Bjørntvedt for the translations.

The derivatives are computed with respect to transforms parameters.
This is the first step towards los calibration.

Thanks to Lucian Barbulescu for the translation.