feat: add 4 gated single agents (Rust Refactor, LLM Post-Training, UI Finish-Gate, Data Viz) (#742)

Lands four gated single-agent contributions via a consolidated branch
(avoids the README-conflict cascade of parallel fork PRs). Each passed
the full gate: lint 0/0, originality 0.0%, canonical structure, and a
manual conceptual-dupe check against the existing roster.

- Rust Refactoring Specialist (engineering) — behavior-aware, evidence-based
  Rust refactoring across crates/traits/modules. (#741, @TanasiDesigns)
- LLM Post-Training Engineer (engineering) — SFT/DPO/GRPO/RLVR experiment
  gating, checkpoint integrity, failure classification. Distinct sub-specialty
  from AI Engineer. (#740, @kaining-never-stop)
- UI Finish-Gate Reviewer (design) — anti-generic UI finish gate; catches
  interchangeable UI before ship via evidence + a written design contract.
  Distinct from the testing-division Reality Checker. (#739, @samuelbushi)
- Data Visualization Engineer (engineering) — chart-type selection, perceptually
  honest encodings, colorblind-safe palettes, performant D3/Vega. (#729, @Hotragn)

README roster rows added for all four; Hermes generated count 265 -> 269.
All guards green (divisions/tools/runbooks/hermes-plugin/lint).

Closes #741
Closes #740
Closes #739
Closes #729

Co-authored-by: TanasiDesigns <TanasiDesigns@users.noreply.github.com>
Co-authored-by: kaining-never-stop <kaining-never-stop@users.noreply.github.com>
Co-authored-by: samuelbushi <samuelbushi@users.noreply.github.com>
Co-authored-by: Hotragn <Hotragn@users.noreply.github.com>
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>

integrations/hermes/README.md

@@ -7,7 +7,7 @@ of adding hundreds of generated skills to `skills.external_dirs`. Hermes sees a
 small fixed tool surface at startup, while the complete Agency roster is
 stored on disk in `data/agents.json` and searched/loaded lazily.
 
-Generated agent count: 265
+Generated agent count: 269
 
 ## Tools exposed to Hermes