Make Prefect an Extra of Cascade

[jhamet93]

At the moment, cascade is coupled to prefect even though there is no fundamental reason that requires this. Users should be able to use this library without prefect. Therefore, this change keeps the existing functionality but requires the user to install prefect in their environment if they wish for the code path of automatic inference to be activated.

Therefore this change introduces the following:

• prefect is now an extra
• prefect<2 support is dropped
• A Context is introduced to represent the execution environment in which different implementations can obtain values that may be needed for remote submission of a job.
• Remove top level prefect imports from the common entrypoints

[AsherWright]

Nice

[gracevivi523]

Overall looks great to me, just a question from my side regarding to Context, would a Context for databricks is necessary or not in addition to gcp and prefect context?

[jhamet93]

Overall looks great to me, just a question from my side regarding to Context, would a Context for databricks is necessary or not in addition to gcp and prefect context?

Internally, databricks has only been the destination for an execution, never a source (i.e. cascade decorated executions are not submitted from Databricks).