Add flow visualization API (Mermaid + ASCII rendering)

[dgenio]

Context / Problem

Flows are opaque data structures. Users can't see the pipeline shape, data flow between steps, or where errors occurred without manually inspecting execution_log records. For debugging, documentation, and demos, a visual representation is essential.

Existing issue #46 covers a CLI viz command, but there is no library-level API for generating visualizations programmatically. Users should be able to render flows in their notebooks, docs, or web UIs.

Proposal

Add library-level visualization methods to Flow and ExecutionResult:

# Static flow visualization
print(flow.to_ascii())
#  [double] ──► [add_ten] ──► [format_result]

print(flow.to_mermaid())
# graph LR
#   A[double] --> B[add_ten]
#   B --> C[format_result]

# Execution result overlay
print(result.to_mermaid())
# graph LR
#   A[double ✓ 3ms] --> B[add_ten ✓ 1ms]
#   B --> C[format_result ✗ ERROR]
#   style C fill:#f66

Core capabilities:

1. flow.to_ascii() — simple ASCII pipeline diagram for terminal output
2. flow.to_mermaid() — Mermaid graph syntax for GitHub README, notebooks, and web rendering
3. result.to_mermaid() — Mermaid diagram with execution status overlay (success/failure/timing per step)
4. flow.to_dot() — Graphviz DOT format for high-quality image rendering (stretch goal)
5. Schema annotations — optionally show input/output field names on edges

Acceptance Criteria

• Flow.to_ascii() returns a string with an ASCII pipeline diagram
• Flow.to_mermaid() returns valid Mermaid graph syntax
• ExecutionResult.to_mermaid() overlays success/failure status and timing on each step
• Failed steps are visually distinct (color/style in Mermaid)
• Generated Mermaid renders correctly on GitHub (test by pasting in a .md file)
• Works for linear flows (1 step, 3 steps, 10+ steps)
• Works for empty flows (no steps)
• Optional show_schemas=True adds field names to edges
• No external dependencies required (pure string generation)
• Tests cover: linear flow, single step, empty flow, failed execution, schema annotations

Implementation Notes

• Implement in chainweaver/viz.py and add methods to Flow/ExecutionResult via mixin or direct methods
• Mermaid syntax: graph LR for left-to-right, graph TD for top-to-bottom
• For future DAG flows (Design and implement DAG-based flow model with topological execution #10), the visualization should naturally handle non-linear topologies
• Consider adding to_svg() that calls Mermaid CLI if available (optional, non-blocking)

Dependencies

• Add dry-run / explain mode for flow execution preview #73 (dry-run/explain mode) — explain data can feed into visualization
• Design and implement DAG-based flow model with topological execution #10 (DAG flows) — visualization should be designed to handle DAGs from the start

Tasks

• Create chainweaver/viz.py with ASCII and Mermaid generators
• Add to_ascii() and to_mermaid() methods to Flow
• Add to_mermaid() method to ExecutionResult with status overlay
• Add optional schema annotation support
• Add unit tests in tests/test_viz.py
• Add visualization usage to README
• Verify Mermaid output renders on GitHub

[dgenio]

Cross-reference: related to #46

This issue (#79 — "Visualization helpers — Mermaid & DOT export") and #46 ("CLI chainweaver viz command") both address flow visualization:

• Add flow visualization API (Mermaid + ASCII rendering) #79: Library-level MermaidRenderer and DotRenderer classes with file export
• Add viz CLI command (ASCII + DOT export) #46: CLI command for quick terminal-based visualization

These are complementary — #46's CLI command would likely call #79's renderers. Consider implementing #79 first as the library layer, then #46 as a thin CLI wrapper.

---

Added during backlog review.