feat: auto discover models from common hosting endpoints

[celeria-ai]

This PR implements the automatic discovery of local models from common hosting endpoints such as LM Studio, Llama.cpp, and MLX.

It introduces the following changes:

• A new check_port function to detect open ports.
• The get_locally_available_models function now scans for LM Studio, Llama.cpp, and MLX on their default ports.
• If a local model provider is detected, it attempts to fetch the list of available models from the /v1/models endpoint.
• Discovered models are added to the list of available models with a provider name like lmstudio-openai-like.

This addresses issue #193.