Add fast-path INSERT method `insert_bulk` for SQLAlchemy/pandas/Dask

[amotl]

Hi again,

in the same spirit as https://github.com/crate/crate-pdo/pull/143, this patch unlocks CrateDB bulk operations for efficient batch inserts using pandas and Dask. It will accompany a followup to the Data Processing and Analytics with Dask and CrateDB: A Step-by-Step Tutorial.

The corresponding documentation section can be inspected at Preview: SQLAlchemy: DataFrame operations.

With kind regards, Andreas.

/cc @marijaselakovic, @hammerhead, @hlcianfagna, @proddata, @WalBeh

[amotl]

Regarding the failing software tests, they should not stop you from reviewing this patch.

I will exclude the relevant test matrix slots from being executed, and add a corresponding .. note:: section to the documentation, about which versions of Python and SQLAlchemy are supported.

[amotl]

Hi.

Thanks for all the excellent review comments.

[This patch] will accompany a followup to the Data Processing and Analytics with Dask and CrateDB: A Step-by-Step Tutorial by @marijaselakovic.

Initially, I rushed a bit to bring this in in time for this tutorial, but I missed the deadline. Now, that we have a little more headroom, I think this patch should only be the foundation, and those aspects should be addressed by corresponding followup patches:

• Expand support for fast-path bulk insert operations to the other DML operations UPDATE and DELETE offered by the CrateDB bulk operations endpoint, when this is sensible within a pandas/Dask environment.
• Don't stop at to_sql(), also cover optimal read_sql() techniques within the documentation, when possible.

I hope you agree with that strategy.

With kind regards, Andreas.

P.S.: Added as backlog items at https://github.com/crate/sqlalchemy-cratedb/issues/74.