We still don't handle all preconnect cases properly

In #1965 we fixed the case for preconnect.

But as we see in https://perfht.ml/30lCQZi there are some cases we still don't handle properly (hover the first request to see the tooltip). I believe this is when there's a preconnect, but that the request starts before the preconnect session. In the code we explicitely force a "one section" layout in this case, but the result isn't correct here obviously.

┆Issue is synchronized with this [Jira Task](https://mozilla-hub.atlassian.net/browse/FP-153)