feat: build the footer row lazily when createFooterRow is enabled at runtime

[6pac-ai]

⚠️ Merge order: this PR assumes #1266 and #1267 are merged first (and pairs naturally with #1268). It has no textual conflicts with them, but it was authored and validated against a master that includes them — please land those before this one.

The bug (Q2 in discussion #1247)

Footer-row DOM was built only in the init path, and internal_setOptions never created it — so setOptions({ createFooterRow: true }) on a live grid flowed into setColumns → createColumnFooter, dereferenced the undefined _footerRowL, and threw, leaving the grid with a half-mutated options state.

The change (the 'support it' resolution from the triage)

internal_setOptions now materializes the footer DOM when the flag flips true. materializeFooterRow() is the single construction path shared with init — the init path's inline footer block is replaced by a call to the same method (R-before-L scroller order, spacers, columns containers, hidden when !showFooterRow), so the two paths cannot drift. On an already-initialized grid it also binds the footer contextmenu/click/scroll handlers; during init those stay bound in finishInitialization, exactly as before. In the runtime path it runs before setScroller (which selects the footer scroll container) and before setColumns (which populates footer cells), so the existing pipeline completes the job with no further special-casing.

Runtime disable hides the footer rather than destroying it — symmetric with showFooterRow — and setFooterRowVisibility works on a runtime-built footer exactly as on an init-built one.

Test

cypress/e2e/quirk-runtime-footer-enable.cy.ts — self-hosting (harness via cy.intercept; no example-page dependency). Enable on a live grid → no throw, scrollers visible, one footer cell and one onFooterRowCellRendered per column, getFooterRow() returns the element, setFooterRowVisibility(false) hides it. Verified to fail pre-fix (TypeError: Cannot read properties of undefined) and pass with the fix; footer + composite-editor + frozen suites ran 27/27 as a regression gate.

One review note: an earlier draft asserted setOptions({ showFooterRow: false }) hides the footer — dropped because master never supported visibility toggling via setOptions for init-built footers either; setFooterRowVisibility is the supported API and is what the spec exercises.

⚠️ Temporary example page

examples/example-quirk-runtime-footer-enable.html is a human-review repro (enable + toggle buttons with a live readout) intended to be deleted before merge — the cypress test is fully independent of it.

(Quirks-triage remaining-items wave: see discussion #1247; siblings #1266, #1267, #1268.)

[ghiscoding]

replicated on my side and all my tests passed, so it seems valid and ready to merge

[6pac]

Actually, I thought it was a bit much repeating all of the footer materialising code for the after init case.
I asked Claude to refactor it to use materializeFooterRow() in the init case as well.

[ghiscoding]

ok that's a bit weird that in that case it ended up with even more loc after the refactor. A quick look at it, looks ok though

[6pac]

From what I can see the refactor dropped 17 lines from slick.grid.ts (+51 -20 after vs +48 before). May have increased the tests LOC?

Anyway, wiil merge!