Use a unified way of handling building the prefix-tree aswell as lookup

[burbas]

This commit makes it so we always are operating on binary strings which makes it much easier to have the same behaviour in lookup and insert.

Things left to do:

• Compare the new and old way in regards to insert

Profiling of insert

Inserting 10 000 random paths with segment length=20 and number of segments=20 (/segment/segment/segment/...) results in the following runtime:

Old solution: ~1.6secs
New solution: ~3secs

This means that the new solution is a bit slower when it comes to inserting new routes. But since this is something that will only happen once or so I think that the upsides with this PR (Unified handling of insert/lookup) outweighs the slowdown.

[Taure]

As you say we do it once. If we need to make it faster we can look at it then.