Address Fluent-centric attribute labels

```
>>> pprint.pprint(solver_session.setup.models.get_attrs(['active?'], recursive=True))
{'attrs': {'active?': True},
'group_children': {'acoustics': {'attrs': {'active?': True},
'group_children': {'model': {'attrs': {'active?': True}}}},
'battery': {'attrs': {'active?': True},
'group_children': {'enabled': {'attrs': {'active?': True}}}},
'discrete-phase': {'attrs': {'active?': True},
...
```
I think it's OK to provide this functionality as it potentially avoids a lot of extra traffic, but those names ought to be translated, including the argument in the call. Background: we originally only supported the raw formats like get_attrs('active?') and added the specific wrapper methods like is_active() later. We didn't address the above though, but I think we should now.

```
>>> pprint.pprint(solver_session.setup.models.get_attrs(['is_active'], recursive=True))
{'attrs': {'is_active': True},
'group_children': {'acoustics': {'attrs': {'is_active': True},
'group_children': {'model': {'attrs': {'is_active': True}}}},
'battery': {'attrs': {'is_active': True},
'group_children': {'enabled': {'attrs': {'is_active': True}}}},
'discrete-phase': {'attrs': {'is_active': True},
...
```

Start at solver_session.setup.models.get_attrs, focusing on recursive attribute lookup and the names passed to it. Compare calls using raw names such as 'active?' with wrapper names such as 'is_active'; done means the argument and returned attribute labels are translated consistently throughout the recursive result.

Written by the indexing model from the issue text.