Disallow literal 0 step in slice expressions

[brianschubert]

Follow up to #18063

Using a step of 0 when slicing a sequence is often a runtime error:

>>> ()[::0]
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
ValueError: slice step cannot be zero

This is true for standard library sequence types (tuple, list, str, memoryview, array.array, etc.), as well as for numpy arrays, pandas series, and many other third-party sequence types (including all that use slice.indices internally).

Strictly speaking, creating a slice with a step of zero is possible, similar to how creating a slice with non-integer-like start/stop/step values is possible. However, just like non-integer start/stop/step values, this is usually a mistake (and one that will cause a runtime error) as opposed to an intended API. Mypy currently warns about non-integer start/stop/step values, so also warning about using a step of 0 seems consistent.

This PR is fairly liberal in what it will warn about: it disallows any expressions with a literal-0 type (or unions including a literal 0) in the step position of a slice expression. Depending on how significant the mypy_primer hit is, I may relax this a bit.

[brianschubert]

mypy_primer hits look good. All hits are from tests that assert that using a step of 0 raises at runtime.

[github-actions]

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

pandas (https://github.com/pandas-dev/pandas)
+ pandas/tests/series/accessors/test_list_accessor.py:100: error: Slice step cannot be 0  [misc]
+ pandas/tests/series/accessors/test_list_accessor.py:103: error: Slice step cannot be 0  [misc]
+ pandas/tests/series/accessors/test_list_accessor.py:118: error: Slice step cannot be 0  [misc]
+ pandas/tests/indexing/test_indexing.py:716: error: Slice step cannot be 0  [misc]

sympy (https://github.com/sympy/sympy)
+ sympy/sets/tests/test_fancysets.py:273: error: Slice step cannot be 0  [misc]
+ sympy/sets/tests/test_fancysets.py:279: error: Slice step cannot be 0  [misc]

[brianschubert]

Thinking about this more, we probably want to move mypy away from this sort of hardcoded (and lint-y) checks. Now that slices are generic, I imagine mypy's handling of slice expressions will be seeing other improvements soon, which may supersede this. Marking this as a draft until those changes settle