Improve factory pattern for agent workflows

[jenna-winkler]

Description

The current factory method for agent workflows in Python lacks clarity in naming and structure. In the current implementation, agents are created and configured using AgentFactoryInput, which is passed to the add_agent method of the AgentWorkflow. A few concerns with this:

1. The class AgentFactoryInput is misleading because it suggests it is responsible for creating agents when it only holds the necessary data to configure an agent

2. The add_agent method both creates and adds agents to the workflow which conflates two different responsibilities

3. The current implementation is a 1:1 match with TypeScript - we need to make it more Pythonic

Possible solutions

1. Rename AgentFactoryInput to AgentConfig ?
2. Separate add_agent method out to 2 methods: add + create

[ajbozarth]

See the fix in #476
I've also reopened #385 to make those updates to the other prublic APIs