How to deal with collective number (same morphemes, different order)?

In a number of Turkic languages, the plural suffix can be used with (usually people?) nouns to mean "that and associated ones".

For example, in Kyrgyz, you can say "Азаматтар" to mean Azamat and his friends/relatives/etc.

You can also say досумдар "my friend and associates". This contrasts with досторум "my friends". In terms of morphology, the difference is solely in the order of morphemes:

```
досумдар:дос>{I}м>{L}{A}р
досторум:дос>{L}{A}р>{I}м
```

The "easy" approach to this would be to just allow analyses to use the same tags in different orders:

```
досумдар:дос
досторум:дос
```

However, this presents a problem for structural transfer and potentially other parts of the Apertium pipeline. It's difficult to condition things on the order of tags in structural transfer, and it's difficult to have a non-canonical order of tags in other places.

Another option would be to make a different tag for this. It could potentially be put in the normal spot for number tags, but semantically and morphologically it makes more sense after any possession tags. Something like the following, perhaps:

```
досумдар:дос
досторум:дос
```

If this number gets a separate tag, we might want to update names to use `` instead of `` too?