flex + grid conformance fixtures

Four pages of small cards. Each card is one property or one value, drawn with fixed 240px containers and 40x20 blocks so the geometry is exact, and each carries a PASS/FAIL badge written by the shared harness in fgcheck.js. A failing card also turns its outline red, so a single screenshot of a page is the whole verdict; the banner at the top gives the count and the log at the bottom lists every failed check with got/want.

• flex1.html — display, flex-direction, flex-flow, flex-wrap, justify-content, align-items, align-content, gap/row-gap/column-gap (30 cards)
• flex2.html — order, flex-grow, flex-shrink, flex-basis (auto/0/px/%/content), the flex shorthand, align-self, auto margins, the automatic minimum size, nesting (27 cards)
• grid1.html — track lists (px/fr/%/auto, repeat, minmax, min-content, max-content, auto-fill, auto-fit), rows, implicit tracks, auto-flow incl. dense, line placement, spans, negative lines, named areas, named lines, the grid-template and grid shorthands (25 cards)
• grid2.html — gaps (incl. percentage), justify/align/place-items, justify/align/place-content, justify/align/place-self, auto margins, order, nesting, spans with alignment (35 cards)

With JS off the pages are still the point: every card is a labelled picture of what the property should do, and the .desc line under the title states the expected numbers. With JS on the harness measures in window.onload and turns that into a count.

Expected values are Chrome's: all four report ALL PASS there (84 + 57 + 85 + 79 = 305 checks over 117 cards), so any FAIL in NetSurf is a NetSurf bug, not a wrong expectation.

Serve them over HTTP (a file:// load runs no JS in NetSurf): python -m http.server 8000 in tests/, then open http://10.0.2.2:8000/fgindex.html in the guest.