Description: The configuration for local_model_name gets overwritten in the .json-config, preventing the addition of a local model in the llm2ssh setup. I think this is caused by the missing initialization of local_model_name in the from_dict method of the Config class.

Steps to Reproduce:

1. Install from pip
2. Configure the local URI and set default_model to "local". Set a local_model_name.
3. Observe that the local_model_name configuration gets overwritten.

Actual Behavior: The local_model_name configuration is overwritten, causing the local model setup to fail.

Environment:
OS: macOS
Python Version: 3.11.2
Package Version: llm2ssh 0.3.7
Local LLM: Ollama running Llama3.1

Proposed Solution:
modify the from_dict method of the Config class in config.py to include the local_model_name field. The updated method may look like this:

@classmethod
def from_dict(cls, d: dict) -> "Config":
    return cls(
      default_model = d.get('default_model', 'groq-llama3-70b'),
      openai_api_key = d.get('openai_api_key', ''),
      claude_api_key = d.get('claude_api_key', ''),
      groq_api_key = d.get('groq_api_key', ''),
      cerebras_api_key = d.get('cerebras_api_key', ''),
      local_uri = d.get('local_uri', 'http://localhost:5000/v1'),
      local_api_key = d.get('local_api_key', ''),
      local_model_name = d.get('local_model_name', ''),  # --> Add this line
      temperature = d.get('temperature', 0.2),
      i_like_to_live_dangerously = d.get('i_like_to_live_dangerously', False),
    )