Consider moving router rebuild to after hook_modules_installed

Let's see what breaks. This fixes 🐛 Fatal when entity type doesn't exist (anymore) Active as expected.

Nothing breaks... interesting. This looks like a sensible change to me but not sure. Here be dragons.

I think there is no reason when hook_modules_installed should come after the router rebuild. Note that if anything does use a route in hook_modules_installed the router will be rebuilt if necessary as this change would mean code would be using the lazy_builder route provider.

I think this makes sense, I'm not sure how we'd test this to be honest. I'd expect to not see the effects until contrib test runs go against it since core didn't trigger any failures.

Not sure if trash is unique? But could it not be mimics from a test module?