fix(test): let the jest scripts exit instead of stalling on open handles

[RonenMars]

npm run test:unit passes all 105 suites and then never exits.
It looks like a hang, and it is easy to kill and record as "the suite hangs" when the run has in fact already succeeded.

Cause

A test that constructs its own QueryClient never clears it, so React Query schedules setTimeout(gcTime) for each cached query once it goes unused, and node will not exit while that timer is pending.

Measured on 2026-08-12:

run	tests take	process exits at
test:unit (105 suites)	~20s	never — killed at 120s, still alive
useConversations.test.tsx -t "partial failure"	1.4s	305s — the 5-minute default gcTime, to the second
useConversations.test.tsx -t "retention gcTime"	0.9s	still alive at 400s — useConversation sets a 7-day gcTime (hooks/useConversations.ts:423)
same file, -t "zzz-no-such-test"	0 tests run	6s, clean

The last row rules out module import as the cause: it is a test body that leaves the timer behind.
--detectOpenHandles reports nothing, which is what makes this hard to place from the symptom alone.

Fix

test:ci already passed --forceExit for exactly this reason.
test:unit, test:integration, test:e2e and test:i18n did not — so the four scripts a developer actually reaches for locally were the four that hung, while the one CI runs was fine.

Giving them the same flag is the whole change.
The same 105 suites, 996 tests, now exit in 20s.

Why not clear the client instead

That was the first plan, and measurement killed it.
The stall does not correlate with test-utils/createWrapper, it correlates with suites that build their own client:

suite	createWrapper calls	own QueryClient	exits on its own?
conversation-search-anchor	12	0	yes, 19s
useSessionActions.resume	0	1	no
useConversations	several	1 (7-day gcTime)	no

Clearing the wrapper's client in a global afterEach would therefore have fixed nothing observable.
Fixing the leak properly means changing the 16 suites that construct their own client, which is a much wider diff than the problem justifies today.
docs/troubleshooting.md now records that finding so the next person does not re-run the same experiment, and names the 16 suites as the target if anyone wants the leak gone rather than worked around.

Trade-off

--forceExit masks a genuinely stuck handle rather than surfacing it.
That trade was already accepted for test:ci; this makes the four sibling scripts consistent with it rather than introducing it.
Everything printed above the stall is a real result, so no test signal is lost.

Verification

• Before, on this branch's base: npx jest --ci --runInBand --testPathPattern='__tests__/unit' — 105 suites pass, Jest did not exit one second after the test run has completed., killed at 120s.
• After: npm run test:unit — 105 suites, 996 tests, all pass, exit=0 at 20s.
• No JS/TS files are touched, so the lint-before-commit rule does not apply.