ci: pin @puppeteer/browsers to v3 to fix Chrome install (#919)

## What

Add a pnpm `overrides` entry to pin `@puppeteer/browsers` to v3 even
though our direct `puppeteer` devDep stays on v24. Also reverts the
misdiagnosed `PUPPETEER_SKIP_DOWNLOAD=1` workaround from #912.

## Why

Reported Node-24-only test failures are caused by a known upstream bug:

- `puppeteer@24` → `@puppeteer/browsers@2.x` → `extract-zip@2.0.1`
(unmaintained since 2020).
- On newer Node releases, `extract-zip` silently aborts mid-stream
without rejecting, exits with code 0, and leaves the Chrome cache
directory existing-but-empty (only the first archive entry, `ABOUT`,
ends up on disk). Subsequent test runs fail with `Could not find Chrome
(ver. 146.0.7680.153)`.

Confirmed empirically — diagnostic instrumentation in
[#916](#916) captured a
clean repro:

```
puppeteer:browsers:install Downloading binary from https://storage.googleapis.com/...
puppeteer:browsers:install Duration for download: 1321ms
puppeteer:browsers:install Installing .../146.0.7680.153-chrome-linux64.zip to .../linux-146.0.7680.153
+ find /home/runner/.cache/puppeteer -maxdepth 4 -type f
/home/runner/.cache/puppeteer/chrome/linux-146.0.7680.153/chrome-linux64/ABOUT     ← only ABOUT
/home/runner/.cache/puppeteer/chrome/146.0.7680.153-chrome-linux64.zip
```

No `Duration for extract` debug line — `unpackArchive()` never resolved.

### Upstream

- Tracking issue:
[puppeteer/puppeteer#14957](puppeteer/puppeteer#14957)
- Fix:
[puppeteer/puppeteer#14960](puppeteer/puppeteer#14960)
(merged 2026-05-11) — replaces `extract-zip` with native
`unzip`/`tar.exe`. Released as `@puppeteer/browsers@3.0.0` (and bundled
into `puppeteer@25.0.0`).

### Why override instead of bumping puppeteer to v25

`puppeteer@25` is ESM-only and requires Node ≥22.12. Both are
surmountable for us, but pinning just the buggy transitive package is a
smaller change with less risk surface:

- This client only uses puppeteer as a Chrome installer for
browser-bundle tests. `test/_helper.ts` calls into `@crawlee/puppeteer`
(CJS, bundles its own `puppeteer-core`); the rest of the test code is
type-only imports. No direct runtime use of puppeteer's main module.
- puppeteer@24's CLI binary is CJS that does
`require('@puppeteer/browsers')`. v3 is ESM-only, but `require()`-of-ESM
is stable on Node ≥20.20 / ≥22.12 / ≥24 — empirically verified on all
three locally with full extraction.
- Lockfile diff: 270 insertions / 490 deletions vs. ~600 insertions /
~1100 deletions for the puppeteer v25 bump.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

---------

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Author: B4nan

.github/workflows/check.yaml

@@ -41,16 +41,6 @@ jobs:
 
             - name: Install pnpm and dependencies
               uses: apify/actions/pnpm-install@v1.1.2
-              env:
-                  # Puppeteer's postinstall kicks off the Chrome download in a
-                  # fire-and-forget Promise (downloadBrowsers() is not awaited
-                  # in puppeteer's install.mjs). When pnpm install returns
-                  # before the download finishes, ~/.cache/puppeteer is left
-                  # half-populated, and the subsequent `browsers install`
-                  # silently no-ops because the directory already exists.
-                  # Skip the postinstall download entirely and let the explicit
-                  # step below do it synchronously.
-                  PUPPETEER_SKIP_DOWNLOAD: 1
 
             - name: Install Chrome for puppeteer
               run: pnpm exec puppeteer browsers install chrome