FIX Auto-wrap synchronous token providers for AsyncOpenAI compatibility

[Copilot]

Users attempting to use get_bearer_token_provider(DefaultAzureCredential(), scope) with OpenAI targets encounter TypeError: object str can't be used in 'await' expression. The OpenAI SDK's AsyncOpenAI requires async token providers, but Azure's get_bearer_token_provider returns synchronous callables.

# This fails with cryptic error
from azure.identity import DefaultAzureCredential, get_bearer_token_provider

api_key = get_bearer_token_provider(
    DefaultAzureCredential(), 
    "https://cognitiveservices.azure.com/.default"
)
target = OpenAIChatTarget(endpoint=endpoint, api_key=api_key)  # TypeError at runtime

Changes

• Added _ensure_async_token_provider() in openai_target.py

  • Detects sync callables via asyncio.iscoroutinefunction()
  • Wraps synchronous token providers in async functions
  • Preserves async callables and string API keys unchanged
  • Logs info message when wrapping occurs

• Integrated into OpenAITarget.__init__()

  • Applied to api_key parameter after environment/parameter resolution
  • Transparent to callers

• Updated type hints and docstrings

  • api_key: Optional[str | Callable[[], str | Awaitable[str]]]
  • Clarified both sync and async token providers are supported

• Added test coverage

  • Unit tests for wrapping logic and all token provider types
  • Integration test validating sync provider with live Azure endpoints
  • Updated existing test expecting synchronous behavior

Result

Both sync and async token providers now work transparently:

# Sync provider (auto-wrapped)
from azure.identity import DefaultAzureCredential, get_bearer_token_provider
api_key = get_bearer_token_provider(DefaultAzureCredential(), scope)

# Async provider (unchanged)
from pyrit.auth import get_azure_openai_auth
api_key = get_azure_openai_auth(endpoint)

# Both work with OpenAI targets
target = OpenAIChatTarget(endpoint=endpoint, api_key=api_key)

Original prompt

---

This section details on the original issue you should resolve

<issue_title>BUG Entra auth with synchronous token providers</issue_title>
<issue_description>#### Is your feature request related to a problem? Please describe.
Since we now support auth token providers via the api_key argument just like the openai SDK it is somewhat unintuitive that synchronous ones won't work. There's a cryptic error being shown if you try that.

E.g., people may try the following with any openai target:

api_key=get_bearer_token_provider(DefaultAzureCredential(), "https://cognitiveservices.azure.com/.default")

which for me yields

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
File /workspace/pyrit/prompt_normalizer/prompt_normalizer.py:95, in PromptNormalizer.send_prompt_async(self, seed_group, target, conversation_id, request_converter_configurations, response_converter_configurations, labels, attack_identifier)
     94 try:
---> 95     responses = await target.send_prompt_async(message=request)
     96     self._memory.add_message_to_memory(request=request)

File /workspace/pyrit/prompt_target/common/utils.py:56, in limit_requests_per_minute.<locals>.set_max_rpm(*args, **kwargs)
     54     await asyncio.sleep(60 / rpm)
---> 56 return await func(*args, **kwargs)

File ~/.conda/envs/pyrit-release-test1/lib/python3.12/site-packages/tenacity/asyncio/__init__.py:189, in AsyncRetrying.wraps.<locals>.async_wrapped(*args, **kwargs)
    188 async_wrapped.statistics = copy.statistics  # type: ignore[attr-defined]
--> 189 return await copy(fn, *args, **kwargs)

File ~/.conda/envs/pyrit-release-test1/lib/python3.12/site-packages/tenacity/asyncio/__init__.py:111, in AsyncRetrying.__call__(self, fn, *args, **kwargs)
    110 while True:
--> 111     do = await self.iter(retry_state=retry_state)
    112     if isinstance(do, DoAttempt):

File ~/.conda/envs/pyrit-release-test1/lib/python3.12/site-packages/tenacity/asyncio/__init__.py:153, in AsyncRetrying.iter(self, retry_state)
    152 for action in self.iter_state.actions:
--> 153     result = await action(retry_state)
    154 return result

File ~/.conda/envs/pyrit-release-test1/lib/python3.12/site-packages/tenacity/_utils.py:99, in wrap_to_async_func.<locals>.inner(*args, **kwargs)
     98 async def inner(*args: typing.Any, **kwargs: typing.Any) -> typing.Any:
---> 99     return call(*args, **kwargs)

File ~/.conda/envs/pyrit-release-test1/lib/python3.12/site-packages/tenacity/__init__.py:400, in BaseRetrying._post_retry_check_actions.<locals>.<lambda>(rs)
    399 if not (self.iter_state.is_explicit_retry or self.iter_state.retry_run_result):
--> 400     self._add_action_func(lambda rs: rs.outcome.result())
    401     return

File ~/.conda/envs/pyrit-release-test1/lib/python3.12/concurrent/futures/_base.py:449, in Future.result(self, timeout)
    448 elif self._state == FINISHED:
--> 449     return self.__get_result()
    451 self._condition.wait(timeout)

File ~/.conda/envs/pyrit-release-test1/lib/python3.12/concurrent/futures/_base.py:401, in Future.__get_result(self)
    400 try:
--> 401     raise self._exception
    402 finally:
    403     # Break a reference cycle with the exception in self._exception

File ~/.conda/envs/pyrit-release-test1/lib/python3.12/site-packages/tenacity/asyncio/__init__.py:114, in AsyncRetrying.__call__(self, fn, *args, **kwargs)
    113 try:
--> 114     result = await fn(*args, **kwargs)
    115 except BaseException:  # noqa: B902

File /workspace/pyrit/prompt_target/openai/openai_chat_target.py:197, in OpenAIChatTarget.send_prompt_async(self, message)
    196 # Use unified error handling - automatically detects ChatCompletion and validates
--> 197 response = await self._handle_openai_request(
    198     api_call=lambda: self._async_client.chat.completions.create(**body),
    199     request=message,
    200 )
    201 return [response]

File /workspace/pyrit/prompt_target/openai/openai_target.py:372, in OpenAITarget._handle_openai_request(self, api_call, request)
    370 try:
    371     # Execute the API call
--> 372     response = await api_call()
    374     # Extract MessagePiece for validation and construction (most targets use single piece)

File ~/.conda/envs/pyrit-release-test1/lib/python3.12/site-packages/openai/resources/chat/completions/completions.py:2678, in AsyncCompletions.create(self, messages, model, audio, frequency_penalty, function_call, functions, logit_bias, logprobs, max_completion_tokens, max_tokens, metadata, modalities, n, parallel_tool_calls, prediction, presence_penalty, prompt_cache_key, prompt_cache_retention, reasoning_effort, response_format, safety_identifier, seed, service_tier, stop, store, stream, stream_options, temperature, tool_choice, tools, top_logprobs, top_p, user, verbosity, web_search_options, extra_headers, extra_query, extra_body, timeout)
   2677 validate_response_format(respo...

</details>



<!-- START COPILOT CODING AGENT SUFFIX -->

- Fixes Azure/PyRIT#1252

<!-- START COPILOT CODING AGENT TIPS -->
---

💬 We'd love your input! Share your thoughts on Copilot coding agent in our [2 minute survey](https://gh.io/copilot-coding-agent-survey).

[romanlutz]

need to validate with integration test locally