Add xdist parallel test task

[tconley1428]

Summary

• add pytest-xdist and change poe test task to use -n auto --dist=worksteal
• make the custom pytest success-exit hook skip os._exit(0) when running under xdist
• switch the main CI test jobs to the parallel task and document the new local command

Validation

• poe build-develop
• poe test-parallel tests/test_common.py
• poe test-parallel tests/test_client.py -k 'test_start_delay or test_query or test_workflow_failed'
• poe test-parallel tests/testing/test_workflow.py
• poe test-parallel tests/testing/test_workflow.py --workflow-environment time-skipping
• poe test-parallel tests/worker/test_workflow.py -k 'hello or multi_param or workflow_info or history_info or signal_and_query or async_utils or simple_activity or simple_local_activity or cancel_activity or simple_child or simple_cancel or cancel_before_run or uncaught_cancel or cancel_child_started or signal_child or cancel_external or signal_external or cancel_multi or cancel_unsent or activity_timeout or with_codec or passthrough_codec or memo_decoding_with_passthrough_codec or custom_runner or continue_as_new'
• poe test-parallel tests/worker/test_workflow.py -k 'typed_config or local_activity_backoff or patch or patch_memoized or uuid or activity_callable_class or activity_method or wait_condition_timeout or typed_handle or memo or custom_failure_converter or optional_param or manual_result_type or dynamic or update_handlers_happy or update_handlers_unhappy or update_task_fails or update_before_worker_start or update_separate_handle or timeout_support or current_build_id_appropriately_set or as_completed_utility or wait_utility or current_update or sleep or concurrent_sleeps_use_proper_options or in_workflow_util or priorities or previous_run_failure or random_seed_functionality'

Timing

• serial first workflow slice: 31.83s
• parallel first workflow slice: 13.09s

[THardy98]

Sweet