Cockpit Command Center: introduce AgentFlowGoalModel, health score, and next-best-action

[smota]

Background & Problem Statement

Epic: #147

Cockpit currently builds views directly from GitHub issue data. This keeps GitHub artifacts as the mental model and makes it hard to express goals, confidence, health, and next best action.

Proposed Solution

Add an intermediate between GitHub source data and UI view models.

Requirements

• Define goal fields: id, title, goalType, status, confidence, nextBestActions, roleContributions, evidenceHealth, highlights, graph, durableLinks.
• Infer from issues, labels, body sections, comments, PRs, checks, replay events.
• Derive explainable next-best-actions.
• Derive confidence/health without over-penalizing missing optional metadata.
• GitHub terminology remains secondary metadata.

Acceptance criteria

• Goal model module exists with tests.
• Next-best-action rules are deterministic and explainable.
• Confidence/health derivation returns score and dimensions.
• Existing board/issue views can consume goal model without GitHub-first naming.

Test plan

• Unit tests for minimal GitHub issue, managed issue, PR-linked goal, blocked review, missing validation, merged goal.
• Regression tests for not rendering ordinary missing optional data as error.

Workflow classification

Architect/developer foundation issue.

[github-actions]

Integrated into development by PR #155: #155

Merge commit: aefe4d9f7d159ccf04d29a4490c4b1c40269e81c
Release/promotion to main is tracked separately.