fix(prompts): 🐛 prompt attention behaviors, add tests

[joshistoast]

Summary

Fixed a case where selecting and adjusting b and c in the prompt (a b)+ c would result in a mangled output, it now correctly adjusts to ((a b)+ c)[attention] instead of (a (b)[attention].

Added tests to ensure any changes follow desired behavior.

Related Issues / Discussions

Automatic Prompt Weighting Behavior Discussion

QA Instructions

Merge Plan

Checklist

• The PR has a short but descriptive title, suitable for a changelog
• Tests added / updated (if applicable)
• ❗Changes to a redux slice have a corresponding migration
• Documentation added / updated (if applicable)
• Updated What's New copy (if doing a release after this PR)

[lstein]

The PR is still marked draft. Is it ready for review?

[lstein]

The use case described in the description is working properly. However, I found a case that does not do what I'd expect:

1. Original prompt: pre (a b c)+ d post
2. Select two or more words within the parentheses: pre (a b c)+ d post
3. Hit the upweight key, get: pre (a b c)++ d post. I expected: pre (a (b c)+)+ d post

Maybe the desired behavior is that any up/down-weighting action within a parentheses expands to the entire contents of the parentheses, but then there is a UI inconsistency, because if I select a single word within the parentheses, or even just have the insertion point adjacent to a word, then that single word is weighted:

1. Original prompt: pre (a b c)+ d post
2. Select one word within the parentheses: pre (a b c)+ d post
3. Hit the upweight key, get: pre (a b+ c)+ d post, which is what I expect

[lstein]

I suggest making the weighting behavior when a phrase is selected consistent with the behavior when a single word is selected.

[lstein]

I've tried multiple scenarios, including changing weighting after selecting across heavily nested parentheses, and could not break it. Nice work!