Bug: [lexical-y] The write self-check reports a shadow root's line where the reader shows inline content in a paragraph of its own #9656

Lexical version: main at 4c04461 (@lexical/y 0.53.0, not released yet), with #9654.

Steps To Reproduce

1. A shadow-root quote holds a paragraph, and next to it inline content that another writer (a plain Yjs client) put there, not in a paragraph: here text inside a link run that also covers the paragraph.
2. Every reader shows that inline content in a paragraph of its own ($inShadowRoot), which is not stored.
3. A local write touches the quote's line without rewriting it (a delete of the quote before it).

With the development self-check on (@lexical/y's test harness enables it), the write reports the quote's line: self-check: line N: text " RAW" != "". The editor's content (two blocks: the shown paragraph and the stored one) is compared with the stored inline content (text, then the stored paragraph).

Link to code example: packages/lexical-y/src/__tests__/unit/regressions/Review9654ReadSafety.test.ts in #9654 (the mixed-before and mixed-after cells, pinned as this report).

The current behavior

The self-check reports a mismatch for a line whose shown content includes a paragraph a shadow root shows around unstored inline content. What clients show and what the write stores are right (the same cells check both).

The expected behavior

The self-check compares such a paragraph's inline content with the stored inline content, as the reader shows it, and reports nothing.