experiment: mark Mapping and Sequence with disjoint_base

[randolf-scholz]

The intersection type Mapping[str, T] & Sequence[T] is non-empty, which can be really annoying for parsing nested data structures. This is an experiment to see what happens when we mark them with @disjoint_base.

[github-actions]

Diff from mypy_primer, showing the effect of this PR on open source code:

Tanjun (https://github.com/FasterSpeeding/Tanjun)
+ tanjun/schedules.py:584: error: Subclass of "Sequence[int]" and "float" cannot exist: have distinct disjoint bases  [unreachable]
+ tanjun/schedules.py:585: error: Statement is unreachable  [unreachable]

mkdocs (https://github.com/mkdocs/mkdocs)
+ mkdocs/utils/templates.py:47: error: Subclass of "ExtraScriptValue" and "str" cannot exist: have distinct disjoint bases  [unreachable]

xarray (https://github.com/pydata/xarray)
+ xarray/core/dataset.py:202: error: Class "Dataset" has incompatible disjoint bases  [misc]

[JelleZijlstra]

• The Tanjun one is an error branch that seems inconsistent with the local annotations (https://github.com/Cursed-Solutions/Tanjun/blob/88d43a267a5bcd995d4929f62b14fbe5c18e59de/tanjun/schedules.py#L584)
• The mkdocs one also looks like incorrect type annotations, I don't think they want multiple inheritance between str and ExtraScriptValue (https://github.com/mkdocs/mkdocs/blob/2862536793b3c67d9d83c33e0dd6d50a791928f8/mkdocs/utils/templates.py#L47).
• But the xarray thing (https://github.com/pydata/xarray/blob/fa2c0ca9a2702d84d82421ea986cc7e8020397a2/xarray/core/dataset.py#L202) genuinely attempts to be both a Sequence and a Mapping.

Overall I think it's worth looking for ways to keep type checkers from considering multiple inheritance scenarios that are very unlikely in practice. Incorrect typeshed stubs might not be the best way though.