Fix idle-resume visibility listener churn

[ejsmith]

Summary

• register document visibility tracking once for each DocumentVisibility lifecycle instead of creating child effects from the visible getter
• keep visibility state reactive for the WebSocket and Intercom consumers
• add an Intercom regression test covering repeated hidden/resume cycles and token refreshes

Root cause

Reading DocumentVisibility.visible from reactive effects dynamically created another child effect and event listener. When visibility or Intercom boot options changed, those child effects were torn down and recreated. The regression test reproduced 502 visibilitychange registrations across 100 hidden/resume cycles; the updated implementation keeps one registration. This listener/effect churn could eventually trip Svelte's effect_update_depth_exceeded guard and leave the page unresponsive after resuming an idle tab.

Validation

• npm run validate
• npm run test:unit -- --run — 299 tests passed
• focused Intercom/WebSocket tests — 31 tests passed
• npm run build
• local frontend and proxied API health checks returned HTTP 200

Breaking changes

None.

[ejsmith]

/preview

[github-actions]

Preview deployed

• App: https://dev-app.exceptionless.io
• Collector: https://dev-collector.exceptionless.io
• Commit: b2cba95b7d85
• Version: 8.6.3-preview.issue.idle-resume-lockup.0.29
• Workflow run: view logs

[github-actions]

Package	Line Rate	Branch Rate	Complexity	Health
Exceptionless.Core	73%	65%	9499	➖
Exceptionless.AppHost	38%	40%	139	❌
Exceptionless.Insulation	23%	23%	205	❌
Exceptionless.Web	84%	67%	6933	✔
Summary	76% (22519 / 29459)	65% (10344 / 15858)	16776	✔