Self-embeddings of models of arithmetic; fixed points, small submodels, and extendability