Enhanced Mode prompt length differs between REST API (20,000 chars) and Python SDK (4,096 chars)

## **Summary**
The Enhanced Mode prompt length limit documented for the REST API is **20,000 characters**,
while the Python SDK documentation states a limit of **4,096 characters**.
This discrepancy causes confusion for users expecting consistent behavior across SDK and REST API.

## **REST API Documentation**
REST API documentation states that Enhanced Mode supports **up to 20,000 characters** for the prompt:

- [REST API documentation](https://learn.microsoft.com/en-us/azure/ai-services/speech-service/llm-speech?tabs=new-foundry%2Clinux%2Cwindows&pivots=programming-language-rest)

## **Python SDK Documentation**
Python SDK documentation states that the prompt limit is **4,096 characters**:

- [Python SDK documentation](https://learn.microsoft.com/en-us/azure/ai-services/speech-service/llm-speech?tabs=new-foundry%2Clinux%2Cwindows&pivots=programming-language-python)

## **Expected Behavior**
The Python SDK should either:
- support the same **20,000-character limit** as the REST API,
**or**
- clearly document why the SDK imposes a stricter limit.

## **Actual Behavior**
- REST API allows 20,000 characters.
- Python SDK documentation states 4,096 characters.
- This inconsistency makes it unclear which limit is authoritative for SDK users.

## **Questions for SDK Team**
1. Should the Python SDK support the REST API’s 20,000-character limit?
2. If not, what is the reason for the stricter SDK limit?
3. Should the SDK documentation be updated to clarify this discrepancy?

## **Environment**
- Package: `azure-ai-transcription`

Comienza comparando la documentación enlazada de la REST API y del Python SDK con el comportamiento del paquete azure-ai-transcription. Confirma qué límite de prompt aplica el SDK y si la diferencia es intencionada; se considerará terminado cuando el comportamiento y la documentación del SDK estén alineados, o cuando el límite más estricto esté explicado claramente.

Escrito por el modelo de indexación a partir del texto del issue.