fix(ui-components): cancel SSMode's deferred listener timer on destroy

[Shivansh1205]

Fixes #1034

Problem

toggleSSMode defers attaching the document click / touchstart listeners by 1ms, working around the activating click firing them immediately. The timer id was never stored, so ngOnDestroy had no way to cancel it.

Destroy the component inside that window and the order is:

1. ngOnDestroy calls removeEventListener — no-ops, nothing is attached yet.
2. The timer fires and calls addEventListener for both events.

The listeners are now owned by a destroyed component, and neither removal path can reach them: ngOnDestroy has already run, and toggleSSMode's else-branch is unreachable once the component is gone. They stay on document for the life of the page, and since the handler is () => document.exitFullscreen?.(), every later click anywhere in the app drops the user out of fullscreen.

This is a half-applied fix

MakePictureComponent has the identical workaround and was hardened for exactly this in ef594ae — "store setTimeout id and clear it in ngOnDestroy to prevent re-adding listeners after destroy" — from review feedback on #867. SSModeComponent was never updated, so that fix covered one of the two copies. This PR is the other one, and deliberately mirrors make-picture's shape so the two stay in step.

Fix

• Store the timer id in ssTimeoutId, cleared inside the callback when it fires normally.
• Cancel it in ngOnDestroy, before the removeEventListener calls.
• Also cancel it in the toggle-off branch, so turning screenshot mode off while the timer is still pending does not attach listeners a moment later.

The shared clearSSTimeout() helper keeps the null-check in one place.

Tests

• does not attach listeners after the component is destroyed — arms the timer, destroys the fixture, runs timers, and asserts nothing was attached for the component's handler afterwards. Fails before this change.
• still attaches listeners on a normal toggle — guards against over-correcting into never attaching them. Passes before and after.

A note on how these are written, since it cost me a wrong turn: onDocumentClick is an arrow-function class field, so it is a stable per-instance reference and removeEventListener genuinely matches it. A test that merely counts add-vs-remove calls therefore nets to zero and passes even against the bug. The assertion has to be about ordering — that an attach happens after ngOnDestroy — which is what these tests check.

I also dropped an assertion on jest.getTimerCount(); zone.js keeps its own timers pending in a TestBed fixture (it returned 6), so it is not a meaningful signal here.

Verification

• Full phoenix-ng suite: 62 suites / 223 tests passed, 0 failures.
• eslint clean on the ss-mode directory.
• No pre-existing failures found, so none are carried here.

[github-actions]

🚀 Preview deployed: http://phoenix-pr-1035.surge.sh

Built from 454ab44.