build(deps): bump pnpm/setup from 2 to 3 in /.github/actions/setup-workspace

.github/actions/setup-workspace/action.yml

@@ -1,4 +1,4 @@
-# Set up workspace: pnpm + Node.js via pnpm/setup@v2, then
+# Set up workspace: pnpm + Node.js via pnpm/setup@v3, then
 # `pnpm install --frozen-lockfile`, each with a bounded retry.
 #
 # Why this exists (#576, flake table): on 2026-09-04 `pnpm/setup@v2` failed
@@ -7,11 +7,11 @@
 # Gateway Timeout` / `fetch failed` — and every one was green on rerun. The
 # action downloads pnpm's executable from the npm registry as its very first
 # network call and exposes no retry input (its inputs are `version`, `dest`,
-# `runtime`, `cache`, `cache-dependency-path`, `working-directory`,
-# `package-json-file`, `require-lockfile`, `install`, `token`), so the retry
-# lives here instead of in every job:
+# `runtime`, `node-version-file`, `cache`, `cache-dependency-path`,
+# `working-directory`, `package-json-file`, `require-lockfile`, `install`,
+# `token`), so the retry lives here instead of in every job:
 #
-#   1. `pnpm/setup@v2` with `continue-on-error`; on failure, wait 15 s and run
+#   1. `pnpm/setup@v3` with `continue-on-error`; on failure, wait 15 s and run
 #      it once more. The second attempt is not `continue-on-error`, so a
 #      second failure fails the job. (`steps.<id>.outcome` is the result
 #      before `continue-on-error` is applied, so it reads `failure` for a
@@ -47,13 +47,13 @@
 # `actions/checkout` must run before it.
 name: Set up workspace
 description: >-
-  Install pnpm and Node.js with pnpm/setup@v2 (retried once on failure), run
+  Install pnpm and Node.js with pnpm/setup@v3 (retried once on failure), run
   `pnpm install --frozen-lockfile` (up to three attempts), and optionally install a
   Playwright browser.
 
 inputs:
   node-version:
-    description: Node.js version for pnpm/setup@v2's `runtime` input, e.g. `22.19.0`, `24`, `26`.
+    description: Node.js version for pnpm/setup@v3's `runtime` input, e.g. `22.19.0`, `24`, `26`.
     required: true
   playwright-browser:
     description: >-
@@ -68,7 +68,7 @@ runs:
   steps:
     - name: Set up pnpm and Node.js
       id: setup
-      uses: pnpm/setup@v2
+      uses: pnpm/setup@v3
       continue-on-error: true
       with:
         cache: true
@@ -80,12 +80,12 @@ runs:
       shell: bash
       run: |
         set -euo pipefail
-        echo "::warning::pnpm/setup@v2 failed on the first attempt (registry 5xx or network error?); retrying in 15 s."
+        echo "::warning::pnpm/setup@v3 failed on the first attempt (registry 5xx or network error?); retrying in 15 s."
         sleep 15
 
     - name: Set up pnpm and Node.js (retry)
       if: steps.setup.outcome == 'failure'
-      uses: pnpm/setup@v2
+      uses: pnpm/setup@v3
       with:
         cache: true
         install: false

docs/local-ci.md

@@ -246,7 +246,7 @@ then treat a repeat as a real signal.
 
 - **Registry 5xx while setting up the job.** Every hosted job installs pnpm,
   Node, and dependencies through the shared `.github/actions/setup-workspace`
-  action. `pnpm/setup@v2` downloads the pnpm executable from the npm registry
+  action. `pnpm/setup@v3` downloads the pnpm executable from the npm registry
   as its first network call and exposes no retry input, so the action retries
   it once after 15 s, then runs `pnpm install --frozen-lockfile` up to three
   times (10 s, then 20 s back-off). A persistent outage still fails the job