fix(imodel-browser): prevent AbortController from breaking fetch in React 18 Strict Mode

[hl662]

Why

Fixes #128.

React 18 Strict Mode double-invokes effects in development: mount -> cleanup -> mount. The previous implementation stored AbortController in useState and had a dedicated cleanup effect:

useEffect(() => () => abortController?.abort(), [abortController]);

On Strict Mode's simulated unmount, this cleanup fired and aborted the in-flight fetch. Because needsUpdate was already set to false inside the effect, no re-fetch was triggered. The hook was permanently stuck at status: Fetching.

As a small boy-scout cleanup, useIModelData.ts also switches from the React namespace import to named hook imports (useCallback, useEffect, etc.) since it is not a component file.
Also, bump node version and pnpm patch

What changed

useIModelData.ts

• Moved AbortController from useState -> useRef (abortControllerRef). A ref mutation doesn't schedule a re-render, so it can't cause the stale-closure / infinite-loop problems that motivated the useState approach.
• Added activeRequestRef = useRef<symbol>(). Each fetch run assigns a new Symbol() as the active request ID. The .then() / .catch() callbacks compare against this ref before applying results -- stale fetches from aborted effect runs are silently dropped.
• Removed the standalone useEffect(() => () => abort(), [abortController]) cleanup.
• Added an unmount-only cleanup effect ([] deps) that clears activeRequestRef and aborts any in-flight request on true component unmount.
• Removed abortController, iModels, and morePagesAvailable from the main effect's dependency array (they were either causing spurious re-runs or are now unnecessary).
• Switched to named React hook imports instead of React.useXxx namespace calls.

useIModelData.test.ts

• Added regression test: "completes fetch when mounted inside React.StrictMode" -- uses { wrapper: StrictMode } and asserts status === Complete with the correct iModel data.
• Placed the test before the it.each cases that spy on window.fetch to avoid test isolation issues with spy leakage.
• Changed afterEach to call jest.restoreAllMocks() alongside server.resetHandlers().

rush.json / ci.yml

• Added ^24.11.0 to nodeSupportedVersionRange per the iTwin.js supported platforms policy (Node 24 LTS supported as of 24.11.0).
• Updated pnpm from 7.32.2 to 7.33.7 (latest pnpm 7.x).
• Updated CI matrix node-version from 20.x to 24.x.

How it works

Strict Mode sequence (before fix):       Strict Mode sequence (after fix):
--------------------------------------   ----------------------------------------
1. Effect runs -> fetch starts           1. Effect runs -> fetch starts (requestId=A)
2. Cleanup fires -> abort() called       2. Cleanup fires -> activeRequestRef = undefined
3. fetch rejects with AbortError         3. fetch resolves -> requestId A != undefined -> dropped
4. needsUpdate=false -> no re-fetch      4. Effect re-runs -> fetch starts (requestId=B)
5. status stuck at Fetching [FAIL]       5. fetch resolves -> requestId B matches -> [PASS]