Update SparseDtype user guide doc

[dsaxton]

Tiny doc nit. The doc says you can pass both dtype and fill_value to SparseDtype then only passes one in the example.

[MarcoGorelli]

The user guide already shows an example on the next line of passing a fill value though - perhaps, instead of making both examples pass a fill value, the wording could be changed so that the first example shows construction with just the dtype of the non-sparse values. Something like

A :class:`SparseDtype` may be constructed by passing the former .. ipython:: python pd.SparseDtype(np.dtype('datetime64[ns]')) , in which case the default fill value for a given NumPy dtype is the "missing" value for that dtype. Note that it may be overridden by passing a `fill_value`, for example .. ipython:: python pd.SparseDtype(np.dtype('datetime64[ns]'), fill_value=pd.Timestamp('2017-01-01')) 

[dsaxton]

Good point @MarcoGorelli. Actually I changed some other wording as well (it's not entirely true that a missing value indicator will be used for numpy dtypes, e.g., not for integers)

In [1]: import numpy as np In [2]: import pandas as pd In [3]: pd.SparseDtype(np.int64) Out[3]: Sparse[int64, 0]

[mroeschke]

Thanks @dsaxton. The CI failure was unrelated

Skipping uploading of coverage data. =================================== ERRORS ==================================== _____ ERROR at setup of TestParquetPyArrow.test_s3_roundtrip_explicit_fs ______