We keep normalizedRequest with data like headers, urls etc. on the sdkProcessingMetadata on the scope.

While this generally works, there are some potential pitfalls/footguns there:

One, if you put some (e.g. partial) normalizedRequest on the current scope, it will just overwrite the full normalizedRequest from the isolation scope, where generally we'll try to put the request data automatically (e.g. in the http instrumentation). Think this example:

Sentry.withScope(scope => {
  scope.setSdkProcessingMetadata({ normalizedRequest: { headers: moreSpecificHeaders } });
});

the resulting event inside of this withScope callback would only have the headers, but would miss e.g. url etc. data set.

This PR changes this so that the normalizedRequest is merged between the types of scopes, so only set fields on e.g. the current scope will overwrite the same fields on the isolation scope, instead of just overwriting the whole normalizedRequest that results.

Note that bundle size for browser is sadly anyhow affected (no matter if we go with a/b/c), as the code to merge it between scopes is always shared :(