Eliminate flex as the GLSL lexical analyzer, going from two nested lexical...

Eliminate flex as the GLSL lexical analyzer, going from two nested lexical analyzers down to one, leaving just the preprocessor's lexical analysis.  A new layer replaces it, to translate from the preprocessor's view of tokenization to glslang's view of tokenization.

Also:
 - change source locations from an int to TSourceLoc (shader number, line number) throughout
 - various improvements to the preprocessor


git-svn-id: https://cvs.khronos.org/svn/repos/ogl/trunk/ecosystem/public/sdk/tools/glslang@22277 e7fa87d3-cd2b-0410-9028-fcbf551c1848