Docs - Add Python SDK API reference documentation

[jenna-winkler]

Description

Currently, the agentstack-sdk Python package lacks a published SDK reference, which means agents (and developers) trying to use the SDK must install the library locally and search through source files to discover correct import paths and usage patterns. The current documentation is scattered and there's no single SDK reference. This creates friction during development and makes it especially difficult for AI coding assistants to reliably generate correct SDK usage.

Proposed solution

Publish a Python SDK API reference similar to https://reference.langchain.com/python/langgraph-sdk

[jjeliga]

@jenna-winkler I have created a first version of the autogenerated Python SDK reference, but I have encountered two issues that currently make it difficult to use in practice.

1. The public export structure is highly inconsistent. Sometimes everything is nicely defined, but in other cases classes are imported directly from seemingly random scripts. Until this is rectified, it is impossible to use any end-to-end reference generation tool. Even when defining the structure semi-manually, it is difficult to determine what should actually be included in the reference.
2. What makes an SDK reference truly usable are docstrings and small examples (usually derived directly from docstrings when they are structured correctly). Currently, there is almost none of that in our SDK codebase, which results in the SDK reference having almost no informational value, except for providing a rough overview of the structure. Until we fill this gap, there is very little value in having an SDK reference.

I propose to postpone the implementation until the state of our codebase allows us to create a valuable sdk reference.