Fix tree names handling

[padovan]

After looking into some of the corner cases for the tree names work, we realized that we need to modify the handling a bit.

Let's change the algorithm to not overwrite every single tree name. Essentially there are two main things we need to fix:

• trees that do not have a tree-name
• trees that appears with different names in multiple origins

Spec:

• It is still okay to pre-populate the file with the Maestro trees first (even if they are not duplicate trees).
• We should also add trees that don't have a tree name to this file in the same way we do today.
• The code should fetch all trees from all origins and look for the duplicate ones across origins.
• A tree is considered duplicated only if it appears in different origins. If the same tree appears in the same origin with different names that is expected (Red Hat does it)
• Then we make that any duplicated trees across origins get an entry on that file.

The goal is to have this file to arbitrate tree names and be the source of truth for the dashboard.

The new /checkout shortcut url should base its url resolution first on this file, and then if it doesn't find anything, it tries to fetch that tree name from the DB. In the example from Red Hat, it will allow to differentiate between different checkouts when the git_url, branch and checkout when only the tree name is different between then.