[BUG] Prompt never properly triggers invocation of managed agent

**Describe the bug**
I have a simple multi-agent setup as in the example (see below).

No matter what question I ask it I can never get the model to trigger the managed agent.
I am using the `o3-mini` reasoning model from Azure OpenAI.

**Code to reproduce the error**
```python
from openinference.instrumentation.smolagents import SmolagentsInstrumentor
from phoenix.otel import register
from smolagents import (
AzureOpenAIServerModel,
CodeAgent,
DuckDuckGoSearchTool,
ToolCallingAgent,
VisitWebpageTool,
)

register()
SmolagentsInstrumentor().instrument(skip_dep_check=True)

model = AzureOpenAIServerModel(
model_id="o3-mini",
api_version="2024-12-01-preview",
reasoning_effort="low",
max_completion_tokens=100000,
)

search_agent = ToolCallingAgent(
tools=[DuckDuckGoSearchTool(), VisitWebpageTool()],
model=model,
name="search_agent",
description="This is an agent that can do web search.",
)

manager_agent = ToolCallingAgent(
tools=[],
model=model,
managed_agents=[search_agent],
)
manager_agent.run(
"What is the current record of the New York Knicks NBA team? Feel free to search the web for the answer."
)
```

**Additional context**
Strangely if I copy/paste the same system prompt and user prompt into the same exact Azure OpenAI model, it behaves as expected


Start by running the provided reproduction with AzureOpenAIServerModel, o3-mini, and the manager_agent.run call, then inspect the ToolCallingAgent managed_agents path and the generated prompts. Compare the prompt and tool invocation behavior with the direct Azure OpenAI example. Done means the manager invokes search_agent for the Knicks question and the regression is covered by an appropriate test.

Written by the indexing model from the issue text.