- Tighten bokeh requirement to match distributed.

- Update to 2023.5.1
* This release drops support for Python 3.8. As of this release
Dask supports Python 3.9, 3.10, and 3.11.
## Enhancements
* Drop Python 3.8 support (GH#10295) Thomas Grainger
* Change Dask Bag partitioning scheme to improve cluster
saturation (GH#10294) Jacob Tomlinson
* Generalize dd.to_datetime for GPU-backed collections, introduce
get_meta_library utility (GH#9881) Charles Blackmon-Luca
* Add na_action to DataFrame.map (GH#10305) Patrick Hoefler
* Raise TypeError in DataFrame.nsmallest and DataFrame.nlargest
when columns is not given (GH#10301) Patrick Hoefler
* Improve sizeof for pd.MultiIndex (GH#10230) Patrick Hoefler
* Support duplicated columns in a bunch of DataFrame methods
(GH#10261) Patrick Hoefler
* Add numeric_only support to DataFrame.idxmin and
DataFrame.idxmax (GH#10253) Patrick Hoefler
* Implement numeric_only support for DataFrame.quantile
(GH#10259) Patrick Hoefler
* Add support for numeric_only=False in DataFrame.std (GH#10251)
Patrick Hoefler
* Implement numeric_only=False for GroupBy.cumprod and
GroupBy.cumsum (GH#10262) Patrick Hoefler
* Implement numeric_only for skew and kurtosis (GH#10258) Patrick
Hoefler
* mask and where should accept a callable (GH#10289) Irina Truong
* Fix conversion from Categorical to pa.dictionary in
read_parquet (GH#10285) Patrick Hoefler