Probably the simplest way would be to have the same effect as

* <translate><!--T:1--> Foo</translate>
* <translate><!--T:2--> Bar</translate>
* <translate><!--T:3--> Baz</translate>

but done automatically:

<translate>
* <!--T:1--> Foo
* <!--T:2--> Bar
* <!--T:3--> Baz
</translate>

That would mean doing three things: always create a new unit when a line starts with list markup (like is currently done for \n\n), always treating them as inline, and removing list markup from the beginning of those units. That would certainly be possible to implement with a bit of work, and I would find it quite useful.

The difficult thing would be figuring out how to make the UX work. There are some situations where it might be useful to not split a list. And translation administrators wouldn't want to always have to split existing lists if they re-tag a page. That means at least adding a new <translate> tag attribute and probably a syntax version too unless the attribute was opt-in instead of opt-out (like nowrap).