When setting up relative URLs for a submodule, then we resolve it to
the actual location and write that into ".git/config" instead of
writing the relative value. We do not yet have a test to nail down this
behaviour, which is now being added by this commit.

In addition to mapping enums to strings in the configuration, we need to
know to delete the configuration option when given the "none" or "no"
option.

The current code will always fail, but only because it's asking for a
string on a live config. Take a snapshot and make sure we fail with
ENOTFOUND instead of any old error.

We no longer have any setters which affect an instance, so
`git_submodule_save()` is no longer relevant.

With this one, we can get rid of the edit_and_save test.

Similarly to the other ones. In this test we copy over testing
`RECURSE_YES` which shows an error in our handling of the `YES` variant
which we may have to port to the rest.

Moving on with the removal of runtime-changing variables, the update
setting for a remote is whatever it was when it was looked up.

Instead of affecting a particular instance, make it change the
configuration.

Having this cache and giving them out goes against our multithreading
guarantees and it makes it impossible to use submodules in a
multi-threaded environment, as any thread can ask for a refresh which
may reallocate some string in the submodule struct which we've accessed
in a different one via a getter.

This makes the submodules behave more like remotes, where each object is
created upon request and not shared except explicitly by the user. This
means that some tests won't pass yet, as they assume they can affect the
submodule objects in the cache and that will affect later operations.