fix(html): make code-tools toggle work when code-copy is enabled

[mcanouil]

The code-copy scaffold introduced in 1.8.15 wraps each code block in a div.code-copy-outer-scaffold, which broke the "Show All Code" / "Hide All Code" handler's direct-child selectors. The menu rendered but clicks did nothing. Selectors now key off details.code-fold and div.sourceCode.cell-code, which the scaffold cannot displace.

• Fixes the fold toggle, which matched zero elements and also mistook the scaffold for the <details>.
• Fixes the hidden/unhidden toggle for echo: false cells under keep-hidden: true, broken the same way but not reported (found/verified during Claude's review).
• Applies the same selector to the server-side toggle detection, which previously worked only because that post-processor happens to run before the scaffold is inserted.
• Also picks up folded blocks with filename, and folded blocks inside tabsets, callouts and listings, which the old direct-child selector never reached.

Verified in headless Chromium by clicking both menu items on a rendered document: all folded blocks open and close together and the hidden block flips both ways. Patching the selectors back to their previous form on the same page reproduces the inert behaviour.

Closes #13583.

[posit-snyk-bot]

✅ Snyk checks have passed. No issues have been found so far.

Status	Scan Engine	Critical	High	Medium	Low	Total (0)
✅	Open Source Security	0	0	0	0	0 issues
✅	Licenses	0	0	0	0	0 issues

💻 Catch issues earlier using the plugins for VS Code, JetBrains IDEs, Visual Studio, and Eclipse.

[cderv]

Thanks @mcanouil.

One thing before merge: we're missing a test that actually exercises the toggle. The smoke fixture locks the DOM shape, but since this PR doesn't touch the render pipeline, the pre-fix version produces identical HTML — so the fixture passes with or without the fix. I confirmed it by reverting the two changed files to their pre-fix state and re-running the new fixture still passes green, so it wouldn't have caught the original regression.

The behavior is pure client-side JS, so it needs a Playwright test that clicks Show All Code / Hide All Code and asserts the <details> open/close and the hidden↔unhidden swap on hidden cells, with code-copy enabled. I'll add that before merging.

[cderv]

Follow-up: added the Playwright test and pushed it to the branch.

tests/integration/playwright/tests/html-code-tools-toggle.spec.ts plus an {ojs} fixture — {ojs} so it needs no Python/R kernel while still producing the same folded + code-copy scaffold DOM. It clicks Show All Code / Hide All Code and asserts every <details> opens/closes and the echo: false cell's source swaps hidden↔unhidden.

It's a real regression guard: it fails on the pre-fix code (the <details> never opens after Show All Code) and passes on this PR. I kept your smoke fixture alongside it — the two are complementary, one locks the rendered DOM contract, the other the JS behavior.

[mcanouil]

Thanks, not yet used to playwright testing.

[cderv]

No problem. 🤖 should know how to write them with the context present. We don't need always some playwright, but for things that impact JS usually we need to verify if the regression can be tested without in-browser behavior or not. Some times our Deno test framework can't test.

I'll merge.

Thanks again !

[mcanouil]

Claude might have caught this, unfortunately, I did it without (quite limited subscription at the moment)