Mark data-diff as type-annotated

The type annotations are incomplete and sometimes simply wrong. But at least they are annotated _somehow_, partially. More importantly, we have numerous simple classes & `attrs` classes, which could be used for type-checking properly without inferring them as `Any` — e.g., in structural pattern matching, such as `match col: case NumericType(precision=0): …`.

Proper type annotations and strict-mode MyPy checking will arrive later, eventually (it is a work in progress, just not a priority).

Author: Sergey Vasilyev