Add shim for Databricks 9.1 [databricks]

[jlowe]

Fixes #3370, #3352.

Adds shim code to support the Databricks 9.1 environment.

[jlowe]

Note that the code in the sql-plugin v2 shims is mostly identical with either 311db or 310until320 with the exception of TreeNode since Databricks decided to pull back that change from 3.2. Otherwise I think it mostly adheres to the same style the 311db shim was doing. Open to suggestions on what names/refactoring to use to reduce the code duplication there.

[jlowe]

build

[tgravescs]

we need to get ci builds setup before checking this in, otherwise builds that try to pull this jar will fail.

[jlowe]

build

[jlowe]

@tgravescs @NvTimLiu please take another look. If this looks correct then I'll revert the changes to the dist pom (so we don't try to pull in the 312db classifier jar when we don't have a Jenkins build setup yet to produce it) and get that through CI. Then we can setup the Jenkins build to produce the spark312db jar and followup with a PR to restore the dist pom changes to pull it into the monolithic dist jar.

[jlowe]

build

[jlowe]

I've manually verified all tests pass, including the cudf_udf tests once the environment is setup (which is a bit different since conda is no longer provided on the node). I left out the changes to dist/pom.xml so this shouldn't be trying to pull in the spark312db shim in the nightly build until we have a chance to setup the nightly CI pipeline to build and publish the artifacts.