feat(ai): agentic information foraging & scent tracker (`devops ai research forage`)

[dan-petty]

Context & Rationale

Implements Pirolli & Card's Information Foraging Theory. Rather than executing disjointed keyword grep searches, the agent follows relational "information scent" across codebases and repositories.

Scope & Deliverables

• Context & Rationale*: Implements Pirolli & Card's Information Foraging Theory. Rather than executing disjointed keyword grep searches, the agent follows relational "information scent" across codebases and repositories.
• Breadcrumb Relational Crawler*: Traverses outbound cues from a starting anchor (error traceback, failing test, or feature keyword) across symbol imports, call hierarchies, git blame history, PR discussions, and markdown references.
• Heuristic Scent Scoring & Backtracking*: Scores edges based on semantic relevance to the inquiry; maintains an explicit traversal ledger that detects circular paths and systematically backtracks when an information trail grows cold.

Strategic Alignment

• Milestone: v0.2.21 (Deep Cognitive Information Foraging, Syntopical Reading & Epistemic Research Engine)
• Roadmap Reference: docs/ROADMAP.md