[Core Feature] Flytekit local caching

[wild-endeavor]

Motivation: Why do you think this is important?
Flytekit local caching

Goal: What should the final outcome look like, ideally?
Unclear, but probably a local file or some other state holding mechanism that flytekit can interact with between python processes that will respect output caching with the same semantics as data catalog. Should also offer an easy way for users to reset this state. Maybe print its contents.

Describe alternatives you've considered
None.

[katrogan]

Additional context:

There are currently 3 execution modes. TASK_EXECUTION is used for tasks in workflows run on hosted Flyte.

A potential first step to implementing this issue might be to take on LOCAL_WORKFLOW_EXECUTION, that is, when a workflow is run end to end on a user's machine. Starting here simplifies implementation since this already handles converting inputs and outputs from python native types to Flyte literals. The local caching story could make use of the existing caching api, specifically: mirroring the CreateDataset and GetDataset interfaces.

Once LOCAL_WORKFLOW_EXECUTION is complete it might be easier to then move on to LOCAL_TASK_EXECUTION for running tasks locally.

[CodeMySky]

I supports this, I think it lowers the barrier on new adopters to test caching locally more easily.

[kumare3]

@CodeMySky this has been checked in, you can use the pre-release or try it out with next Flyte release