Allow tool async to sync conversion

[chenmoneygithub]

Because DSPy evaluator and optimizer doesn't support async yet, we provide a backdoor to execute async tools in sync mode so that optimizer and evaluator can still work.

We make a specific flag allow_tool_async_sync_conversion to enable the conversion to avoid unintended conversion happens for the user.

With this PR the following code works:

import asyncio

import dspy
from dspy.datasets import HotPotQA

dspy.configure(lm=dspy.LM("openai/gpt-4o-mini"), allow_tool_async_sync_conversion=True)


async def search_wikipedia(query: str) -> list[str]:
    await asyncio.sleep(0.01)
    results = dspy.ColBERTv2(url="http://20.102.90.50:2017/wiki17_abstracts")(query, k=3)
    return [x["text"] for x in results]


trainset = [x.with_inputs("question") for x in HotPotQA(train_seed=2024, train_size=500).train]
react = dspy.ReAct("question -> answer", tools=[search_wikipedia])

tp = dspy.MIPROv2(metric=dspy.evaluate.answer_exact_match, auto="light", num_threads=24)
optimized_react = tp.compile(react, trainset=trainset)