Dask-sql

Features
* Allow for multiple schemas (205)
* AutoML capabilities (199)
* Implement the `regr count` SQL operator (193)
* ML model improvement : Added `SHOW MODELS`, `EXPORT MODEL` and `DESCRIBE MODEL` (185, 191)
* Implement the `search` and `sargs` operator (184)

Bugfixes
* Fixes for pandas 1.3.0 (202)
* Fix test fixture order (194)
* Fix a failing build, as ciso8601 is currently not pip-installable (192)

Bugfixes
* Casting of literals is done by Calcite, except for strings (178)

Bugfixes
* Speed up aggregations when there are no aggregates (174)
* Register the lower and upper-case version of a function (177)
* Reverting a bug in the casting logic to cast only if really needed (176)

Small feature addons
* Added correct casting and mod operation (172)
* Implement `OVER` for arbitrary windows (164)
* Allow to start a SQL server from a jupyter notebook (162)

Bugfixes and Improvements
* Sort optimizations (167, 173)
* Fix scikit learn version in docker file
* Add test with independent dask cluster (165)
* Speed up builds with mamba (171)
* Remove version constraints for pandas and dask as the errors were fixed upstream (170)
* Fixed the replacement of functions/aggregations and added a test (169)
* Added missing version in pom

Small feature addons
* Allow function reregistration (161)
* upgrade fugue dependency (160)
* Implement a wrapper for the prompt_toolkit session (159)

Small feature addons
* First working (but slow) implementation of `OVER` (157)
* Add a visualize function (153)
* IPython/Jupyter Magic (146)
* Hive/Databricks from SQL (145)

Bugfixes and Improvements
* Improve documentation
* Better cross joins (150)
* Fix a bug which occurs when only filters are present in groupbys (154)
* Make testing a bit easier to type
* Fix a warning on regexes
* Split out the jupyter notebook integration (152)
* Add pre commit hook (149)
* Limit the dask version until the dask-ml problem is fixed (147)
* Turn off docker image building of PRs
* Fix integration with dbfs using the newest fsspec version (140)
* Show a reasonable traceback on exceptions (142)
* Docker image improvements (137)
* Support for Float (pandas extension type) and filter with NaNs (136)