docs(readme): add a data-driven loop program graph beside the linear one

[abrichr]

What

The launcher README shows the compiled MockMed workflow as a linear program graph. This adds a second, richer example right after it -- a data-driven loop -- so the README shows both the straight-line case and the control flow a linear IR cannot express.

The graph is a compact save-encounter body wrapped by openadapt flow for-each into a governed LOOP, rendered by openadapt flow visualize --format mermaid:

flowchart TD
  n0{"Loop over worklist"}
  n1("type patient_id")
  n2("type note")
  n3("save encounter<br/><small>effect · irreversible</small>")
  n4{{"Success"}}
  n0 -->|per row of worklist| n1
  n1 --> n2
  n2 --> n3
  n3 -->|next record| n0
  n0 -->|worklist exhausted| n4

Loading

It shows the Loop over worklist node running the body once per record (per row of worklist), the next record loop-back edge, the worklist exhausted exit to Success, and the irreversible, effect-verified save encounter write as the per-record halt point. A plain-text caption carries the same content for PyPI (which does not render Mermaid), matching the linear example's style.

Honesty

• The graph is engine-generated, not hand-drawn, and the loop it depicts is replay-tested in openadapt-flow (a companion PR expands loop bodies in visualize and ships the committed docs/showcase-encounter-loop bundle the runtime interprets once per record).
• Identity gating is stated as "where armed" (this body's keypress save shows the effect gate + irreversible halt, not an identity gate); the body is labelled a compact MockMed fixture, with a pointer to the full-recording loop showcase.
• No new claims beyond what the engine produces; explicit if/else branching is not depicted.

Depends on openadapt-flow to render this exact graph. The Mermaid embedded here is static and already correct; no code in this repo changes.

🤖 Generated with Claude Code

https://claude.ai/code/session_01NyCHrzA1psrKMFfroYbzaM