chore(bench): evict ten stale zero-reference experiment scripts

[drewstone]

What

Removes ten one-off experiment scripts from bench/src: david-goliath, david-attribution, atom-humaneval, aec-gate, trata-gate, humaneval-object-ablation, hev-improve, hev-eval, diverse-gate, decoder-live.

Why

• All ten have zero references anywhere in the repo (rg-verified at this HEAD, both pre- and post-fetch).
• June-dated; several pin dead default models (groq/llama-3.1-8b-instant, gpt-4o-mini) and hand-rolled price tables superseded by agent-eval's native cost accounting.
• agent-runtime keeps substrate + benchmark adapters; experiments belong in agent-lab (its AGENTS.md law 1). The David-Goliath lineage is re-implemented fresh there: agent-lab#419.

Proof

• bench tsc error count identical before/after deletion (3 pre-existing workspace-resolution errors, untouched files) — eviction is typecheck-neutral.
• No test file references any evicted script.

[tangletools]

✅ Auto-approved drewstone PR — e4817fca

This PR was opened by the trusted drewstone account.
The full PR reviewer audit still runs separately and will publish findings if it detects issues.

This approval is provisional. It rests on the audit running. If the audit cannot run — for example the CLI bridge rejects it — this approval is dismissed rather than left standing, so an unrun check never reads as a passing one.

_{tangletools · auto-approval · reason: drewstone_author · 2026-08-15T06:58:11Z}