Add word-break setting to boundary analysis

[valadaptive]

This is necessary for implementing linebender/parley#303.

In the long term, Parley should move to some other Unicode analysis crate, but this should do for now.

[valadaptive]

Actually, let me see if there's a way to do this per-cluster

[valadaptive]

I've got an implementation that seems to work for Parley, but I suspect may still be incorrect.

The API is not very elegant--the iterator returned by text::analyze now has a set_break_strength method. This makes iteration more awkward if you choose to use it, but has the benefit of not breaking backwards compatibility.

I'm not sure what the extra entries in the pair table are for, but they seem to carry some extra state. For implementing break-all, which requires characters to be treated as ID if they're AL | NU | SA, I'm checking if the previous character and current character are AL | NU | SA, and treating each as ID for the purposes of the current line break opportunity only. I'm not sure how this interacts with the extra pair table entries, which is why I say this may be incorrect.

[xorgy]

This makes sense to me.

[xorgy]

Upon looking more closely at the Parley PR I think it might make sense to consolidate the definitions of WordBreakStrength since they are otherwise the same.

[valadaptive]

I've updated the definition of WordBreakStrength here so it can be used in Parley.