build(eslint): add the missing config, and stop capping eslint's major

[BernardJen]

Why

npm run lint has never worked (#63). The script shipped from the start, but no config ever did — so eslint exited non-zero on every invocation, and nothing caught it because no workflow ran it.

Fixing that also unblocks #51. That PR was failing at npm ci with ERESOLVE, not at any check step — worth stating plainly, because it's easy to assume a "bump eslint" PR fails on lint:

While resolving: eslint-plugin-react@7.37.5
Found: eslint@10.9.0
Could not resolve dependency:
peer eslint@"^3 || ^4 || ^5 || ^6 || ^7 || ^8 || ^9.7" from eslint-plugin-react@7.37.5

eslint-plugin-react has no release that accepts eslint 10 — the next tag is a stale 7.8.0-rc.0 peering ^3 || ^4. It was also dead weight: with no config, nothing had ever referenced it.

What changed

• Dropped eslint-plugin-react, bumped @typescript-eslint/* to ^8.69.0 — the first line peering eslint ^8.57 || ^9 || ^10. Nothing in the tree caps eslint's major any more.
• Added eslint.config.mjs, deliberately not built on @eslint/js, which peers a single eslint major and would recreate this exact failure on the next bump.
• Fixed the lint script — it still passed --ext, removed in eslint 9.
• CI runs lint now, between install and typecheck. A lint script no job runs is how this went unnoticed for the life of the repo.
• Dead code removal to reach a clean run: unused imports, two never-read catch bindings, and currentPresentationPath in both Windows backends — written twice, read never. macos.ts genuinely reads it as a fallback, so it stays there.
• Dependabot grouping (lint-and-types, vite, react).

On dependabot grouping

#51 was one of four PRs failing the same way: a solo bump of a package whose peers are pinned to a sibling's major can never resolve alone. #49 and #50 are the clearest case — one bumps vite to 8, the other bumps @vitejs/plugin-react to a version requiring vite 8, and each fails precisely because the other is a separate PR.

Still blocked upstream — not fixed here

• Bump typescript from 5.9.3 to 7.0.2 #47 (typescript 7) — @typescript-eslint peers typescript >=4.8.4 <6.1.0, and latest 8.69.0 still does.
• Bump @vitejs/plugin-react from 5.1.2 to 6.1.0 #49 / Bump vite from 7.3.5 to 8.2.2 #50 (vite 8) — grouping merges them into one PR, but electron-vite@5.0.0 (latest stable) peers vite ^5 || ^6 || ^7. Only 6.0.0-beta.1 accepts vite 8.

Grouped, both now sit still instead of failing every week.

Known warnings

0 errors, 18 warnings. require-atomic-updates is set to warn rather than disabled — it flags seven pre-existing interleaving hazards in the automation backends, where module-level slide state is read before an await and written after. They look real, but fixing them means reworking concurrency in code only exercisable against a live PowerPoint install. Left visible so the debt stays counted. The other 11 are no-explicit-any at the COM and PowerShell boundaries.

Verification

lint, typecheck, test (7 passed), build and a clean npm ci all pass locally. Separately confirmed that setting eslint to ^10.9.0 on this branch resolves cleanly and lints clean — i.e. #51 will go green once rebased.

Note: #60 will need a dependabot rebase, since package-lock.json moved here.

Closes #63

🤖 Generated with Claude Code