The current script extractors will be rebased onto this, while also
allowing built-in C++ extractors next to them using the same interface.
This will replace the current generic/custom extractor split.