Stack traces without in-app stack frames group all together

### Environment

SaaS (https://sentry.io/)

### Steps to Reproduce

Example: https://dropbox.sentry.io/issues/7470118519/events/0aff6e56b53e4f7b9d005144a03da0c5/?alert_rule_id=14776445&alert_type=issue&environment=alpha¬ification_uuid=2a8883f3-a8c9-451c-b30c-efbef8df0a72&project=4505546600087552

Another example: https://dropbox.sentry.io/issues/5826139225/events/c1c39459b24548eaa8283ef070ad3e31/?project=4505546600087552&query=is:unresolved%20level:%EF%80%8DContains%EF%80%8Dfatal&sort=freq

### Expected Result

We would like these to group better. We'd expect that if in-app frames are not available that non-in-app frames could be used.

### Actual Result

Anything without in-app frames gets grouped together and becomes unactionable, as well as obscuring our ability to access impact per actual problem.

Tab through the events, you'll see they do not appear related, often having totally different stack traces.

### Product Area

Issues

### Link

https://dropbox.sentry.io/issues/5826139225/events/c1c39459b24548eaa8283ef070ad3e31/?project=4505546600087552&query=is:unresolved%20level:%EF%80%8DContains%EF%80%8Dfatal&sort=freq

### DSN

_No response_

### Version

_No response_

Start by tracing the issue-grouping path for events without in-app stack frames, using the two linked Dropbox Sentry examples to compare their distinct traces. Determine where non-in-app frames could affect grouping, then add coverage showing unrelated stack traces no longer collapse into one issue and verify the examples produce actionable groups.

Written by the indexing model from the issue text.