GetGdalReadImageFileName does not set its second argument when its first
argument is a non empty directory and not a supported format (OTB
supports opening directories in some cases). But in that case it does
return false, which was not handled before this commit. This resulted in
a misleading error message and a corrupted m_FileName.

This changeset introduce support for Gdal derived subdatasets.

Changes include:
- A new private const method to get the real source name from the derived dataset filename
- An update of the TestFileExistency() method to check for existency of the real source name instead of derived dataset filename
- A patch to make sure OSSIM tries to open the real source name instead of the derived dataset.

If derived datasets are not supported by gdal, ImageFileReader will end with GDALImageIO CanRead saying false, so there should be no impact.

Used the following command:

find . \( -iname *.txx -and ! -path *ThirdParty* \) -exec perl -pi -w -e 's/__otb(.*)_txx/otb$1_txx/g;' {} \;
find . \( -iname *.h -and ! -path *ThirdParty* \) -exec perl -pi -w -e 's/__otb(.*)_h/otb$1_h/g;' {} \;

Fixes many, but not all, clang -Wreserved-id-macro warnings.

ENH: manage the outputpixelType to the imageIO level to remove include otbGDALImageIO from otbImagefilereader

modifications generated via the script FindRedundantHeaderIncludes.py