refactor(datashare-python): register workers through entry point plugins

[ClemDoum]

Changes

Implement workflows and activities discovery through Python entrypoints for plugin.

When devs create worker packages, and installs it together with datashare-python, they can then start their worker running the workflow and activities of the choice as following:

datashare-python worker start --queue <queue> [--activities <activity-regex>] [--workflows <workflow-regex>] --temporal-address localhost:7233

This API lets the CLI user cherry picking the activities and workflow of their choice and route them to a given queue.

The registry machnism works as following:

1. the dev must expose all workflows and activites in some variable inside their package say in a worker_package/registry.py:

from .workflow_a import WorkflowA
from .workflow_b import WorkflowB
from .activiy_a import actitiy_a_fn
from .activiy_b import ActB

WORKFLOWS = [WorkflowA, WorkflowB]
ACTIVITIES = [actitiy_a_fn, ActB.act_fn]

2. they must then expose these variables [registering an entrypoint](entrypoints for plugin) in their pyproject.toml under the datashare.workflows and datashare.activities plugin/entrypoint group:

[project.entry-points."datashare.workflows"]
workflows = "worker_package.registry:WORKFLOWS"

[project.entry-points."datashare.activities"]
activities = "worker_package.registry:ACTIVITIES"

After that the datashare-python CLI will auto discover registered activities and workflow and perform filtering through regexes

Note: workflows and activite must use explicit name through the name= keywoard arg when using the activity/workflow.defn(name=) function

datashare-python

Added

• workflow and activities registration through plugin entry points