Routing supports dead-end u-turns

(copied from https://github.com/a-b-street/ltn/pull/119#discussion_r1954042416)

>I do not have any counter-examples and will have to think about how to force one, but this rule looks like it prohibits U-turns at a deadend. I don't think a route would ever choose to go down one, unless maybe there's a situation where it's forced to by a TR... ie:
>
>```
> D
> |
> A--*--B
>```
>
> If the route starts from B, goes to *, there's a TR preventing B->A, then the route goes to D. It needs to u-turn there to make progress. Then southbound, D->A.
>
> I don't think something this weird is likely to happen in practice, so don't block this PR on it. I've made a note to try and construct this test case later and see if it can actually occur; can file a followup issue if you agree it's possible

The issue describes a potential routing edge case involving U-turns at dead-ends when turn restrictions force a route. Start by examining the routing algorithm code, likely in a module handling pathfinding or turn restrictions. Look for existing tests related to U-turns or dead-ends. The goal is to construct a test case that reproduces the scenario described (B -> * -> D -> U-turn -> A) and verify if the current algorithm handles it correctly. Understanding the graph representation and turn restriction logic is key.

Written by the indexing model from the issue text.