Initial support for configuring agents in a flow

About

@leetickett-gitlab is in discussion with Agent Platform team around support for accepting flow definitions for AI Catalog, and wrote initial steps for that support in #553147 (comment 2662974877).

Our flow definitions currently support "steps":[{"agent_id": 1}].

It's early in this process, as support for us to send data to Workflow Service is in development in https://gitlab.com/gitlab-org/modelops/applied-ml/code-suggestions/ai-assist/-/issues/1270. However, the experimental Flow YAML format has documentation that lets us peek at what is supported in hard-coded Flows.

Is there a logical starting place for configuring AI Catalog agents within our flows, which can lead to next steps for backend and frontend?

It would be great to be able to map out a plan so we can begin to work on this - if it's possible to at this stage.

Proposal

TBD!