Suggest mode: Reject type-over overlapping a suggestion marker

[i-am-chitti]

What?

In Suggesting mode, typing over a selection that overlaps an existing inline suggestion marker now cleanly rejects the keystroke (with a snackbar explaining why) instead of falling through to native editing, which corrupted the content.

Part of #73411. Fixes #79799.

Why?

The addition/deletion keyboards intentionally decline to build a suggestion when the current selection overlaps an existing marker (wrapping the range in a del would re-attribute part of that marker to a new id). The problem was how the addition keyboard declined: it returned without cancelling the event, so the browser edited a DOM that still contained the <mark class="wp-suggestion">. That native surgery corrupted the content:

• the typed leading space was dropped,
• the marker was fragmented into several <mark> segments, and
• the paragraph occasionally emptied entirely.

This is the pre-existing behaviour documented in #79799 (not a regression from the readEventRange DOM-caret work in #79714 / #79752).

How?

• suggestion-addition-keyboard.js: insertText now returns a tri-state — INSERT_HANDLED (wrote a suggestion), INSERT_REJECTED (deliberately declined but must still cancel native editing), INSERT_DEFERRED (no valid inline anchor — let it fall through). The onBeforeInput / onPaste handlers preventDefault on HANDLED and REJECTED, and only fall through for DEFERRED. The type-over-overlap guard returns REJECTED, so the browser never touches the marker DOM.
• A snackbar ("You can't type over an existing suggestion. Accept or reject it first.") is shown on reject — with a stable notice id so a burst of keystrokes refreshes one notice instead of stacking — so the swallowed keystroke doesn't read as an unresponsive editor.
• Building a well-formed combined deletion+addition suggestion for this case is left to a follow-up phase; this PR guarantees the "cleanly rejected, no content change" outcome from the issue.

Why the deletion/cut guards are left unchanged

They share the same structural shape, but their fall-through is already captured cleanly as a whole-attribute suggestion (the reconciler + store interceptor), pinned by the existing e2e test "seam: a delete straddling an existing marker never corrupts it". Rejecting them would remove that working behaviour, so this PR scopes to the addition type-over that actually corrupts.

Testing Instructions

1. Enable the Suggestion Mode experiment (Gutenberg → Experiments).
2. Add a paragraph Hello world and switch the editor mode to Suggesting.
3. Select world and press Ctrl/Cmd+B — world gets wrapped in a format suggestion marker.
4. With world still selected (do not collapse the caret), type more.

Expected: nothing is inserted, the paragraph still reads Hello world, the world marker stays a single intact <mark> (no dropped space, no fragmented markers, block not emptied), and a snackbar explains why the keystroke did nothing.

Automated: test/e2e/specs/editor/various/suggestion-mode-overlay-retirement.spec.js adds a regression test ("type-over of a selection overlapping a marker is rejected, not corrupted"). Whole spec passes locally (9/9), including the delete-straddle test that proves no regression.

Screenshots or screencast

Before -

After -

Usage of AI

Claude Code - Opus 4.8

[github-actions]

The following accounts have interacted with this PR and/or linked issues. I will continue to update these lists as activity occurs. You can also manually ask me to refresh this list by adding the props-bot label.

If you're merging code through a pull request on GitHub, copy and paste the following into the bottom of the merge commit message.

Co-authored-by: i-am-chitti <iamchitti@git.wordpress.org>

To understand the WordPress project's expectations around crediting contributors, please review the Contributor Attribution page in the Core Handbook.