fix: add validation for unsupported model settings parameters

[hafsatariq18]

Fixes #856

This PR adds validation for model settings parameters to prevent silent failures when invalid or unsupported parameters are provided. Previously, passing invalid parameters like "tempural" would not trigger any validation errors.

Changes:

• Add validate_model_settings function to check for invalid parameters
• Update merge_model_settings to validate settings before merging
• Add comprehensive tests for model settings validation
• Provide clear error messages for invalid parameters

Example usage and error:

model = OpenAIModel(DEPLOYMENT_NAME, openai_client=client)
agent = Agent(model)
result = await agent.run(prompt, result_type=People, model_settings={"tempural": 45})

UserError: Invalid model setting parameter(s): tempural. Valid parameters are: frequency_penalty, logit_bias, max_tokens, parallel_tool_calls, presence_penalty, seed, temperature, timeout, top_p