fix: remove the duplicated main landmark from the root layout

[defimomof2]

What this changes

Removes the duplicated <main> element from the root layout, and adds a unit test so the same breakage cannot reach main a third time.

Why

app/layout.tsx on main currently renders two <main> elements and {children} twice:

<SiteNav />
<main className="container">{children}</main>
<SiteFooter />
<AnalyticsScript />
<WebVitalsReporter />
<main className="container" id="main" tabIndex={-1}>
  {children}
</main>

Two consequences, both user-visible:

• Every page renders its entire content twice.
• The skip link is broken. Its target is #main, which is the second element — so a keyboard user activating "Skip to main content" lands below a full duplicate of the page rather than at the start of the content. That is the opposite of what 2.4.1 asks for.

It came from a merge that resolved a conflict in the layout body by keeping both sides — the analytics branch's <main> and the accessibility branch's <main id="main" tabIndex={-1}>. Keeping the latter preserves the skip-link target; the footer and analytics components move back below it, where they were.

Why nothing caught it

This is the second time this exact block has been broken this way, and every existing gate is blind to it:

• tsc and ESLint are both perfectly happy with two <main> elements.
• landmark-unique is an axe best-practice rule, not WCAG A/AA — and e2e/accessibility.spec.ts runs only wcag2a, wcag2aa, wcag21a, wcag21aa, so the suite stays green.
• Nothing asserts that {children} appears once.

tests/unit/layout-landmarks.test.ts closes that gap at the source, asserting the element count, the {children} count, and the skip-link wiring (href="#main" has a target that carries both id="main" and tabIndex={-1}, and sits ahead of the nav). I confirmed it fails against the current layout rather than passing vacuously.

A source-level test rather than a DOM one on purpose: the failure mode is a merge splicing the file, and that is exactly what this catches — cheaply, without a browser.

How it was verified

npm run lint      ✔ no warnings or errors
npx tsc --noEmit  ✔ clean
npm test          ✔ 113 passed (20 files) — 4 new
npm run build     ✔ exit 0

• Existing tests pass
• New tests cover the change
• No change to a public contract interface

Risk

Very low, and it removes a defect rather than adding behaviour. The only semantic change is that content renders once instead of twice and the skip link reaches the top of it.

[joelpeace48-cell]

Nice job @defimomof2