feat/fix: Align the word break behaviour of Aspiers' new text transformation functions more closely with readlines conventions.

[ariane-emory]

Type of change

• Bug fix
• New feature
• Refactor / code improvement
• Documentation

What does this PR do?

This PR is an enhancement/fix to a PR on the upstream repo (anomalyco#6778), which provides additional readline/Emacs-style text transformations, that aligns the word break behaviour more closely with readline/Emacs conventions, addressing the issue that I pointed, out in my recent comments on the upstream PR.

[aspiers]

Thanks for this! I noticed your PR while looking at the upstream PR — and I had already pushed an equivalent fix to my branch earlier today (see my comment on the upstream PR).\n\nOur implementations are functionally equivalent: both switch from /\s/ (whitespace) to /\w/ (alphanumeric + underscore) for word character detection. Yours preserves the original branching structure and extracts a nice isWordChar helper; mine simplifies the whole function to a single forward-scan pattern. I may adopt the isWordChar helper for readability — it does make the intent clearer.\n\nI'll compare our test suites too to see if yours covers anything mine doesn't.

[aspiers]

I've now pushed a "best of breed" version to my upstream PR that incorporates the best aspects of both of our fixes. Here's a summary of what ended up in the final version:

From your PR:

• The isWordChar helper function — good for readability and avoiding repeated inline regex literals.
• The input_transpose_characters entry added to types.gen.ts — your PR caught that this was missing.

Fixes beyond both our initial attempts:
After verifying against Emacs and Readline directly, I found two further issues that affected both our implementations:

1. _ is not a word character. We both used /\w/ which includes underscore, but Emacs uses its word syntax class (only [A-Za-z0-9]) and Readline uses isalnum(). Neither includes _. So foo_bar is two words, not one.

2. No "fallback to previous word" when at end of buffer. We both fell back to the previous word when the cursor was past the last word. But Emacs upcase-word/downcase-word is simply a no-op in that situation — verified with emacs --batch. The function now returns null and the handler skips the action.

The final isWordChar is /[A-Za-z0-9]/ and getWordBoundariesForTransformation is now a clean two-phase forward scan: skip non-word chars, advance through word chars, return null if no word found.