[lagru]

When releasing v0.7.0, I wasn't sure if numpy.dtype would accept unions, so I implemented them as an expansion and named the grammar rule accordingly.

But now I found examples in NumPy's repo like this one

np.ndarray[ShapeT, np.dtype[np.floating | np.integer | np.bool]],

So, I'm quickly deprecating that unnecessary functionality.

Release note

For maintainers and optionally contributors, please refer to changelist's README on how to document this PR for the release notes.

The grammar rule 'dtype_to_expand' is deprecated in favor of 'dtype_union'.
Since `numpy.dtype` supports unions, the rule was renamed. Use the rule
'dtype_union' if you want to modify this expression in a plugin.
Using the rule in a plugin will trigger a warning and will stop working in v0.8.0.