Optimize SQS task performance by extracting models once

[coderabbitai]

Problem

In agave/tasks/sqs_tasks.py, the run_task function currently extracts request and response models on each execution using get_request_model(task_func) and get_response_model(task_func). Since the task function doesn't change between executions, this is unnecessary and inefficient.

Proposed Solution

Move the model extraction out of the run_task function so it happens only once when the task is defined or initialized, and then reuse these models for each task execution.

For example, this could be done in the task decorator or when creating a new task instance:

# Example implementation
def task(...):
    def task_builder(task_func: Callable):
        # Extract models once
        request_model = get_request_model(task_func)
        response_model = get_response_model(task_func)
        request_log_config_fields = get_sensitive_fields(request_model)
        response_log_config_fields = get_sensitive_fields(response_model)
        
        @wraps(task_func)
        async def start_task(*args, **kwargs) -> None:
            # Pass pre-extracted models to run_task
            ...
            
        return start_task
    
    return task_builder

This optimization would improve performance for frequently executed tasks.

Related PR: #159 (comment)