Currently pydoc supports only limited set of data descriptors: builtin member and getset descriptors (this covers slot descriptors and structseq member descriptors) and properties. But it does not fully support custom data descriptors.

For example, after implementing accelerators for namedtuple fileds access in issue32492, if P = namedtuple('P', 'x y'), help(P.x) will output the help for the _tuplegetter class instead of the P.x member.

The proposed PR replaces checks for particular types of data descriptors with a general check. It performs also some refactoring and adds a bunch of tests.