fix(jobs): pass useRef an explicit initial value

[umutcankurt]

One line of code, but it removes the wall that has been hiding what the React 19
upgrade (#15) actually involves.

The problem

React 19 removed useRef's zero-argument overload, so this fails against its types:

src/pages/JobHistory.tsx(30,32): error TS2554: Expected 1 arguments, but got 0

It is the only argument-less useRef in the codebase — 1 of 8 call sites —
and CI on #15 stops there, at the type-check step, before the tests or the
build ever run. So nobody knew what was behind it.

The change is React 18-safe

Not a forward-port that breaks the present. React 18's third overload is:

function useRef<T = undefined>(initialValue?: undefined): MutableRefObject<T | undefined>;

Passing undefined explicitly resolves to that same overload, so the ref keeps
its exact type. Verified rather than assumed — a scratch probe asserting mutual
assignability between the old and new forms type-checked clean against the
installed React 18 types:

const before = useRef<ReturnType<typeof setTimeout>>();
const after  = useRef<ReturnType<typeof setTimeout>>(undefined);
const _a: typeof before = after;   // ✅
const _b: typeof after = before;   // ✅

What is behind the wall — measured

With this fix applied, React 19.2.8 + @types/react 19.2.17 installed locally
as a throwaway experiment (reverted afterwards; this PR does not bump React):

gate	result under React 19
tsc --noEmit	0 errors
npm run lint (+3 postlint gates)	pass
npx vitest run	458 tests / 34 files pass
npx vite build	pass

Also swept for the APIs React 19 removed, since green gates do not prove runtime
behaviour:

removed / changed in 19	occurrences
ReactDOM.render / hydrate	0
findDOMNode	0
string refs (ref="name")	0
defaultProps on function components	0
propTypes	0
legacy context (childContextTypes)	0
react-test-renderer	0
unmountComponentAtNode	0

The entry point already uses createRoot. forwardRef appears 12 times and is
deprecated-but-supported in 19, so it is cleanup, not a blocker.

Caveat worth stating plainly: all of the above is static and test-level. It
says the React 19 migration is far smaller than feared, not that it is risk-free
— StrictMode and effect-timing changes are exactly the kind of thing this
project's 34 test files would not catch. React 19 still deserves its own PR with
a manual pass over the app.

Verification of this PR itself (React 18, unchanged)

npm run lint, npx tsc --noEmit, npm run i18n:check, npx vitest run
(458 tests / 34 files), npx vite build — all pass.

🤖 Generated with Claude Code