fix: model register missing model-type and not accepting metadata

[Bobbins228]

What does this PR do?

Added the ability to specify the model type when registering models.
Also fixed a bug with passing metadata which would result in the following error:

Error Type: BadRequestError                                                                                                                                     │
│ Details: Error code: 400 - {'error': {'detail': {'errors': [{'loc': ['body', 'metadata'], 'msg': 'Input should be a valid dictionary', 'type': 'dict_type'}]}}}

Closes: #215

Test Plan

Run the following commands

# Note a Llama Stack Server must be running
# Create a venv
uv sync --python 3.12
# Install the LSC with the new code changes
uv pip install -e .
# List the available models
llama-stack-client models list
# Register the granite-embedding-30m embedding model NOTE must have sentence-transformers as an inference provider
llama-stack-client models register granite-embedding-30m --provider-id "sentence-transformers" --provider-model-id ibm-granite/granite-embedding-30m-english --metadata '{"embedding_dimension": 384}' --model-type embedding
# Verify the embedding model added are present
llama-stack-client models list

[mattf]

thank you! this fixes both llamastack/llama-stack-client-python#215 and llamastack/llama-stack-client-python#216

[mattf]

lgtm