Using `setTimeout` in the Router navigation pipeline creates fragile and
unpredictable behavior. Additionally, it creates a new macrotask, which
would trigger an unnecessary change detection in the application.

This `setTimeout` was added in
angular@15e3978.
Both tests added in that commit still pass. Additionally, the comment
for _why_ the `setTimeout` was added doesn't really explain how the
described bug would occur. There has been a lot of work in the Router
since then to stabalize edge case scenarios so it's possible it existed
before but doesn't anymore.

Removing this `setTimeout` revealed tests that
relied on the navigation not completing. For example, the test suite did
not have a route which matched the redirect, but the test passed because
it ended before the redirect was flushed, so the `Router` never threw an
error. Similar situations exist for the other use of `setTimeout` in the Route
(the one in the location change listener).
There were no other failures in TGP other than incorrectly written
tests.

BREAKING CHANGE:
When a guard returns a `UrlTree`, the router would previously schedule
the redirect navigation within a `setTimeout`. This timeout is now removed,
which can result in test failures due to incorrectly written tests.
Tests which perform navigations should ensure that all timeouts are
flushed before making assertions. Tests should ensure they are capable
of handling all redirects from the original navigation.