Add option for output type (uint8, unit16,..., float, double)

[alexbruy]

in OTB tools is possible in (all?) tools to define the bit depth of output, example

-out <string> [pixel] Output Image [pixel=uint8/uint16/int16/uint32/int32/float/double] (default value is float) (mandatory)

by default is "float" that many times is not the best option. As in Processing is not possible to change this default then further manipulation of outputs is often needed.

The problem is that this [pixel] definition is not added to the OTB command as any other parameter: it must be added just after the "-out" parameter without any "-" in front of it, exemple:

otbcli_KMeansClassification -in input.tif -out output.tif uint8

this means that this parameter cannot be added in OTB tools definitions files because they will be added to the command line with the "-" causing an error for unrecognized parameter.

I guess that this means that OTBAlgorithm.py must be patched to cope with the "non" parameter.

Moved from QGIS bugtracker https://issues.qgis.org/issues/11032