simplify JDTBatchCompiler.getUnits(). Reuse existing CU

[pvojtechovsky]

The compilation is done in two phases.

1. treeBuilderCompiler.buildUnits()
2. parser.dietParse() this call is processed for each result of the treeBuilderCompiler.buildUnits() and needs a CompilationUnit with the same name ... it is the same file which was used to produce the unit ... so we can use the CompilationUnit which is directly stored here unit.compilationResult.compilationUnit

I would like to have this change because

1. it makes code more readable
2. it makes clear that same compilation unit is used. Before the treeBuilderCompiler.buildUnits() worked with different input files, then parser.dietParse() ... it took me several hours until I recognized that both inputs should be always same.
3. the parameter files of the method CompilationUnitDeclaration[] getUnits(List<SpoonFile> files) is now unused, so we can remove it in next PR