More efficient route lookup caching

MR does two things:

1. When generating the cache ID, runs the same logic as the decode inbound path processor. That processor should be merged into the router, all it does by trying to decouple it is create inconsistencies. See 🐛 After updating Drupal from version 10.2.7 to version 10.3.0, a 404 error occurred! Needs review for more.

2. When no routes are found, instead of caching a serialized empty route collection, caches FALSE instead. Given 404s are unbounded, while this is a relatively small saving, it's potentially multiplied by a lot of cache items.

Moved the URL normalization stuff into 🐛 After updating Drupal from version 10.2.7 to version 10.3.0, a 404 error occurred! Needs review .

early returns make sense.

Not for this issue but would be neat to eventually have a performance test around caching if possible?

I did think about test coverage here but all we could really do is assert the contents of the cache entry when no routes are found rather than a performance test, which feels a bit too much like repeating what the code does.

nod_ → closed merge request !8795

nod_ → reopened merge request !8795

some phpcs issue

Sorry, sloppiness.

nod_ → closed merge request !8795

Committed and pushed da5f6215a9 to 11.x and 41ac5c1446 to 10.4.x. Thanks!

Automatically closed - issue fixed for 2 weeks with no activity.