feat(adaptive-delivery): add host configuration doctor

[BjRo]

Why

Adaptive Delivery needs a read-only way to explain host delegation failures without accidentally activating orchestration. This addresses the effective-config, concurrency, and nesting diagnosis requested in #161 while aligning the implementation with Darrow's current contained Python mechanics.

What changed

Adds the self-contained doctor-adaptive-delivery skill for Codex and Claude Code and derives baseline and full-path capacity from the documented delegation topology. The deterministic doctor now lives in the existing darrow-adaptive-delivery Python package, uses frozen UV entrypoints and standard TOML parsing with a Python 3.10 backport, supports native Windows configuration discovery, and removes the Bash runtime implementation. The branch also updates the normative invariant, discovery and troubleshooting docs, representative evals, copied-plugin coverage, and both plugin manifests to version 0.23.0.

Verification

bun run check:python passed across every registered package; the changed Adaptive Delivery package passed 205 tests with 98.84% line and 97.55% branch coverage. The fresh copied plugin passed with runtime-only dependencies, the skill inspector and Claude manifest validator passed, and docs, Prettier, ESLint, TypeScript, shell lint, decisions, and diff checks passed. Both-host five-case dry evals prepared successfully; focused live Codex and Claude trials each passed task and activation checks. Repository-wide bun test still contains the unrelated evals/runner/discovery-eval-checks.test.ts planning-transfer failure (expected false, received undefined), reproduced in isolation.

Review notes

The highest-risk code is the host-configuration parser and capacity derivation in backend/src/darrow_adaptive_delivery/doctor.py. Tests cover malformed and wrongly typed TOML, unreadable and absent files, isolated CODEX_HOME, native Windows USERPROFILE, V1/V2 depth behavior, Claude version boundaries and defaults, redaction, complete remediation guidance, command exits, and copied runtime-only installation. Native Windows behavior is covered by the repository's Python matrix; this local verification ran on macOS.

Checklist

• I have read and followed CONTRIBUTING.md, including the contribution
  licensing terms.
• I added or updated the applicable invariant before implementation, or
  this change does not affect a capability invariant.
• I added or updated colocated evals, or this change does not affect skill
  behavior.
• I confirmed that each changed plugin remains self-contained, or this
  change does not affect plugin content.
• I ran bun run check:python, or this change does not affect registered
  Python packages or their repository quality infrastructure.