[otnc]

Follow-up to the monorepo and TypeScript work (#17, #18). Why: npm 10.9 crashes resolving the workspace (edgesOut) so lockfile updates needed npm 11; pnpm is strict about undeclared dependencies, has --filter for per-package scripts, and matches funmary-app.

• pnpm-workspace.yaml, packageManager: pnpm@11.22.0, pnpm-lock.yaml; package-lock.json and the npm workspaces field removed
• Packages declare their own build tools (typescript, tsdown, vitest, @types/node) because pnpm does not hoist undeclared ones; root gets vitest explicitly
• Root scripts use pnpm; typecheck:ts / test:packages use pnpm --filter \"./packages/*\" (quoted so the shell does not expand the glob; the root has its own test/typecheck scripts for MoonBit)
• CI and _release.yml: pnpm/action-setup (v6.1.0, SHA-pinned) before setup-node with cache: pnpm; pnpm install --frozen-lockfile; tests/builds through pnpm --filter; publishing and version bumps still use the npm CLI in each package directory (Trusted Publishing is an npm feature; npm version no longer needs the -w workspace flag). Release commits stage only package.json (and moon.mod for the compiler), since pnpm has no root lockfile entry for versions
• Docs, skill and .prettierignore updated. End users are unaffected (purus new is package-manager agnostic)

Also fixes: dist/ build outputs of the wrapper and both packages were accidentally committed in #85 (that branch predated the dist/ ignore entry); they are untracked in a separate commit.

Verified locally: pnpm install --frozen-lockfile, pnpm run build, typecheck, ESLint, Prettier, wrapper tests (25), package tests (12 + 8), per-package builds; npm version and npm pack in a package directory. The release workflow itself cannot be exercised without publishing; the first real release after merging is the test.

Also removes packages/mcp/dist/index.js: compiled output of the third-party purus-mcp (unlicensed, not part of this repo) that was swept into #72 by a git add -A from leftovers of the abandoned import (packages/mcp and packages/language-server leftover directories were also cleaned locally). The file still exists in the history of #72; rewriting main would need the ruleset lifted, so it is left to the maintainer to decide.