This PR does two things:

1. Removes any usage of propagation context on isolation scope. We never really need/read this, as the current scope one will always overwrite this. Generally we should only ever set/read propagation context on current scope.
2. Adjust server-side withIsolationScope to reset the propagation context unless it is continuing a trace. Today, this kind of works in otel-based places because we wrap root spans in startNewTrace. But this does not align with core/browser handling, so going forward this will be removed. instead, we want this logic to be applied in an isolation scope, so the following code will result in separate traces:

Sentry.withIsolationScope(() => {
  Sentry.startSpan({ name: 'process 1' }, () => ...);
});

Sentry.withIsolationScope(() => {
  Sentry.startSpan({ name: 'process 2' }, () => ...);
});

without this change, both of these spans would be part of the same trace, which is not what we want in server.

note: In browser, we do not actually fork the isolation scope, there is just a single one, so this does not apply.