Fix crash in thread updating

[vkrause]

Happened at startup/sync in Itinerary, due to Thread::addEvent() being called with a nullptr event from here.

[KitsuneRal]

If it's nullptr how did it end up in the timeline?

[KitsuneRal]

What you're guarding against is rather the thread root event somehow is not RoomMessageEvent, and it normally just shouldn't happen. I'm just trying to make sure we're not masking a problem instead of fixing it.

[vkrause]

Right, I have no idea about the context of all this, and it's very well possible this is just hiding the problem (the application isn't doing anything with threads on its own). I can reliably reproduce this in one installation here, so I can test other approaches or diagnostics as well.

[nvrWhere]

Agree with @KitsuneRal here all thread roots should be room message events. That being said I don't think the server would stop a client from adding a thread relation to an invalid root. So I think we need this but it's also worth checking that there isn't something else going on.

@vkrause can you get the JSON for the problem event