Improved FastFlow support in RPLsh

In this thesis I extended RPLsh, a tool for the analysis and generation of parallel programs built via composition of algorithmic skeletons. In particular, I added a new skeleton, the divide and conquer, with new rewriting rules and annotations for non functional parameters; I fixed and redesigned the code generation that now works with the latest release of FastFlow (3.0), and I improved the overall user experience, with better error reporting and a new auto-completion functionality