fix: Stop re-running event processors on every main process scope change

[timfish]

Fixes #1467

Every scope change in the main process rebuilt the event defaults by running every event processor, including the synchronous app.getAppMetrics() (about 0.04 ms per process), and built the merged scope three times over. A burst of console.log calls or forwarded renderer breadcrumbs blocked the main thread for seconds in apps with many processes.

A new createScopeSaver merges the scope changes of one event loop turn into a single save on the next turn, reads the merged scope once when that save runs, and rebuilds the defaults only when they are older than 5 seconds. Both minidump integrations call its refresh from app.whenReady(), because screen and GPU context only exist once the app is ready. A sequence number stops a slow older save from writing over a newer one. onScopeChange in common/scope.ts registers a bare callback on the three scopes; addScopeListener is unchanged for the renderer. The Electron uploader path now builds its event synchronously, so its epoch guard is gone.

Unit tests cover coalescing, the single scope read, caching, refresh, the stale-save guard and rejection handling. The reproduction from the issue now makes zero app.getAppMetrics() calls for 5,000 console.log calls and the main thread is busy for about 18 ms instead of seconds (about 12 ms with no SDK). The full native crash e2e suite passes on Electron 39.8.10.

app.getAppMetrics() still runs once per real event through ElectronContext; whether the app-wide memory figure is worth keeping is a separate question.

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit b78fda8. Configure here.}