[SPARK-48705][PYTHON] Explicitly use worker_main when it starts with pyspark

[HyukjinKwon]

What changes were proposed in this pull request?

This PR proposes to use worker modules when it starts with pyspark.

Why are the changes needed?

Because now daemon uses worker module from the first argument, it could break if users are using custom workers with providing parameters. They aren't actual API but it is better to be more strict on the argument provided.

Here is an example of sentry-python

ModuleNotFoundError                       Traceback (most recent call last)
File <command-558609978944307>, line 4
      1 import sentry_sdk
      2 from sentry_sdk.integrations.spark import SparkWorkerIntegration
----> 4 sentry_sdk.init(
      5     dsn="https://[email protected]/0",
      6     traces_sample_rate=1.0,
      7     profiles_sample_rate=1.0,
      8     integrations=[
      9         SparkWorkerIntegration(),
     10     ],
     11 )

File /.../python3.10/site-packages/sentry_sdk/hub.py:121, in _init(*args, **kwargs)
    115 def _init(*args, **kwargs):
    116     # type: (*Optional[str], **Any) -> ContextManager[Any]
    117     """Initializes the SDK and optionally integrations.
    118 
    119     This takes the same arguments as the client constructor.
    120     """
--> 121     client = Client(*args, **kwargs)  # type: ignore
    122     Hub.current.bind_client(client)
    123     _check_python_deprecations()

File /.../python3.10/site-packages/sentry_sdk/client.py:128, in _Client.__init__(self, *args, **kwargs)
    124 def __init__(self, *args, **kwargs):
    125     # type: (*Any, **Any) -> None
    126     self.options = get_options(*args, **kwargs)  # type: Dict[str, Any]
--> 128     self._init_impl()

File /.../python3.10/site-packages/sentry_sdk/client.py:162, in _Client._init_impl(self)
    155 if self.options["request_bodies"] not in request_bodies:
    156     raise ValueError(
    157         "Invalid value for request_bodies. Must be one of {}".format(
    158             request_bodies
    159         )
    160     )
--> 162 self.integrations = setup_integrations(
    163     self.options["integrations"],
    164     with_defaults=self.options["default_integrations"],
    165     with_auto_enabling_integrations=self.options[
    166         "auto_enabling_integrations"
    167     ],
    168 )
    170 sdk_name = get_sdk_name(list(self.integrations.keys()))
    171 SDK_INFO["name"] = sdk_name

File /.../python3.10/site-packages/sentry_sdk/integrations/__init__.py:124, in setup_integrations(integrations, with_defaults, with_auto_enabling_integrations)
    120 logger.debug(
    121     "Setting up previously not enabled integration %s", identifier
    122 )
    123 try:
--> 124     type(integration).setup_once()
    125 except NotImplementedError:
    126     if getattr(integration, "install", None) is not None:

File /.../python3.10/site-packages/sentry_sdk/integrations/spark/spark_worker.py:31, in SparkWorkerIntegration.setup_once()
     28 @staticmethod
     29 def setup_once():
     30     # type: () -> None
---> 31     import pyspark.daemon as original_daemon
     33     original_daemon.worker_main = _sentry_worker_main

File /.../python/pyspark/daemon.py:40
     37 if len(sys.argv) > 1:
     38     import importlib
---> 40     worker_module = importlib.import_module(sys.argv[1])
     41     worker_main = worker_module.main
     42 else:

File /usr/lib/python3.10/importlib/__init__.py:126, in import_module(name, package)
    124             break
    125         level += 1
--> 126 return _bootstrap._gcd_import(name[level:], package, level)

Does this PR introduce any user-facing change?

Yes, it recovers the support of sentry-python.

How was this patch tested?

Manually tested.

Was this patch authored or co-authored using generative AI tooling?

No.

[HyukjinKwon]

Merged to master.