A framework for creating and curating high-quality code datasets tailored for large language models