The conversion to int16 will be done after orthorectofication

However the output mask receives a 2 instead of 0 (nodata), or 1 (nominally
computed)

And check most likelly invalid states first ({0,0,0} is more likelly to happen
than nan's)

Avoid float -> double -> float conversions: works on floats all the way down

- remove unused includes
- add missing ones (in the right file)
- reorder: OTB > ITK > 3rd part > STD

- harmonize indent and use latest OTB style (Allman's indenting style)
- remove trailling spaces

This commit doesn't need much attention. It serves to avoid extra noises in
futures commits that would automatically change indenting...

and rename the files to avoid conflict

... to invalid normals, DEM values...

... to nodata, NaNs, "folded" neighbours (when left and right are a the same
position...)