Add Python samples for durabletask-python v1.4.0 features

[andystaples]

Summary

Adds 4 new Python sample directories and updates 1 existing sample to cover the major features added in durabletask-python v1.4.0.

New Samples

large-payload/

Demonstrates automatic externalization of oversized orchestration payloads to Azure Blob Storage using BlobPayloadStore. Shows both small (inline) and large (externalized) payloads transparently handled by the SDK. Includes a docker-compose.yml to start the DTS emulator and Azurite together.

work-item-filtering/

Demonstrates running multiple specialized workers (worker_a.py, worker_b.py) where each only processes specific orchestrations/activities using use_work_item_filters(). Shows auto-generated filter mode.

testing/

Demonstrates unit testing orchestrations and activities without Docker or an emulator using InMemoryOrchestrationBackend from durabletask.testing. Includes pytest-based tests with happy path and failure path coverage.

orchestration-management/

Demonstrates client-side lifecycle management operations: restart_orchestration() (same ID and new ID), purge_orchestrations_by() for batch cleanup, and get_all_orchestration_states() for querying.

Updated Sample

async-http-api/

Replaced the synchronous DurableTaskSchedulerClient with AsyncDurableTaskSchedulerClient for native async/await support. Uses azure.identity.aio.DefaultAzureCredential, FastAPI lifespan for client lifecycle, and async context manager pattern. Added documentation section to README.

Conventions

All new samples follow existing conventions:

• Kebab-case folder names
• worker.py / client.py split
• Standard env var handling (TASKHUB, ENDPOINT)
• READMEs following SAMPLE_TEMPLATE.md structure
• requirements.txt with minimal dependencies

Verification

• All .py files pass python -m py_compile syntax validation