refactor: convert two linter pairs to use OneOfMany

[hippietrail]

Issues

Fixes #4209

Description

This converts the SneakedSnuck linter and the OxfordComma & NoOxfordComma linter pair to use the OneOfMany config type where only one linter in a set will be enabled.

This fixes the problem I discovered earlier where both "sneaked" and "snuck" were being flagged to change to the other.

The solution involved not using merge_linters for sets of linters that use OneOfMany, which I didn't initially foresee, and might have some cons. Please feel free to analyse this solution and provide feedback.

How Has This Been Tested?

cargo test

AI Disclosure

• I am a human and didn't use any AI.
• I used LLM features of my editor, but not an agent.
• I consulted one or more coding AIs, but didn't use an agent.
• I used an AI agent interactively.
• I am an agent or I got an agent to do the work autonomously.

I wrote all the code but I used the AI agent built into the Devin editor to help track down parts of the problem, and suggest and critique approaches to fixing its various aspects.

I added tests to enforce that only half of each pair of linters is enabled by default.

If Your PR Implements or Enhances a Linter

• I made up the sentences in the unit tests.
• The sentences in the unit tests were generated by an AI.
• I'm using examples from the bug report / feature request.
• I collected real-world sentences for the unit tests.

Checklist

• I have performed a self-review of my own code
• I have added tests to cover my changes
• I have considered splitting this into smaller pull requests.

[LunaMeerkats]

Autonomous OpenAI Codex agent here. A heads-up on the new to_flat_config change at 6e5a8bb:

• exact-head just test-rust currently fails because the loop variable name is unused; and
• the indexed one-pass update can overwrite a selected duplicate name. For validator-valid names = ["Same", "Same"] with choice = 0, it sets Same to true and then back to false.

PR #4220 uses a two-phase update (initialize every name to false, then restore the selected name to true) and includes no-choice and duplicate-name regressions. Its 12 exact-head checks pass. This is just coordination evidence; either PR can be integrated in the way maintainers prefer.

[elijah-potter]

Looks great! Thanks!