feat(azure): add Responses API support for Azure OpenAI provider

[kunalk16]

Closes: #5202

Summary
Adds support for OpenAI's Responses API to the Azure OpenAI provider. Users can now opt in by setting api="responses" when creating an LLM instance. The existing Chat Completions API path remains the default and is completely untouched.

The Responses API is OpenAI's recommended API for new development, replacing Chat Completions as the primary interface. Azure OpenAI has adopted it with full feature parity.

# Example 1 basic
llm = LLM(model="azure/gpt-5.4-mini", api="responses")

# Example 2 with instructions
llm = LLM(model="azure/gpt-5.4-mini", api="responses", instructions="Be concise.")

Click to view Testing Details

Smoke Testing using the below test script:

from crewai import LLM

API_KEY = "api-key"
ENDPOINT = "https://some-internal-name.openai.azure.com"
MODEL = "azure/gpt-5.4-mini"

llm = LLM(
    model=MODEL,
    api="responses",
    api_key=API_KEY,
    endpoint=ENDPOINT,
)

print("=== Basic call ===")
result = llm.call(messages=[{"role": "user", "content": "Say hello in one sentence."}])
print(result)

print("\n=== Reasoning (medium) ===")
llm_reason = LLM(
    model=MODEL,
    api="responses",
    api_key=API_KEY,
    endpoint=ENDPOINT,
    reasoning_effort="medium",
)
result = llm_reason.call(messages=[{"role": "user", "content": "What is 47 * 83?"}])
print(result)

print("\n=== Multi-turn conversation ===")
llm_chat = LLM(
    model=MODEL,
    api="responses",
    api_key=API_KEY,
    endpoint=ENDPOINT,
    auto_chain=True,
    store=True,
)
r1 = llm_chat.call(messages=[{"role": "user", "content": "My favorite color is blue. Remember this."}])
print(f"Turn 1: {r1}")
r2 = llm_chat.call(messages=[{"role": "user", "content": "What is my favorite color?"}])
print(f"Turn 2: {r2}")
llm_chat.reset_chain()

print("\nDone!")

Test Output: