udecode/plate v53.0.7 on GitHub

`@platejs/ai`

#4987 by @zbeyens – Add transformInitialValue and nodeId.initialValueIds while keeping normalizeInitialValue as a deprecated alias
#4987 by @zbeyens – Improve large-document mount and render performance across core element, mark, and nodeId paths
- Cut 10k mixed-document core mount time from 1240.60 ms to 468.26 ms without nodeId (62.3%, 2.65x faster)
- Cut 10k mixed-document core mount time from 1290.66 ms to 477.73 ms with nodeId (63.0%, 2.70x faster)
- Cut mixed-document nodeId overhead over core from +50.06 ms to +9.46 ms (81.1% smaller)
- Cut duplicate-id paste cost from 20.06 ms to 13.79 ms (31.2%, 1.45x faster)
- Cut 10k code-only mount time from 1500.30 ms to 496.47 ms (66.9%, 3.02x faster) and shrink the code-only tax over core from +280.75 ms to +27.89 ms (90.1% smaller)
- Bring the current 10k core and basic large-document mount lanes to Slate parity or better (core -3.5%, core + nodeId -1.6%, basic -1.2%)
- Preserve Slate children for void render.as tags and Slate attributes on simple leaf/text render paths

Published @platejs/csv@53.0.7.

#4987 by @zbeyens – Add experimental editor.tf.setNodesBatch for exact-path node prop updates on large documents
- Cut the large-document set_node hot path from 18.56 ms to 2.63 ms at 1k blocks (7.05x faster)
- Cut the same path from 118.54 ms to 4.92 ms at 5k blocks (24.10x faster)
- Let @platejs/core batch live nodeId normalization instead of paying one setNodes call per missing id
- Keep editor.tf.setNodesBatch explicitly temporary. It is experimental and will be removed in a future release
231b986 by @github-actions[bot] – Updated slate-hyperscript.

#4987 by @zbeyens – Speed up unmerged table range selection while preserving merged-cell handling

Published @platejs/test-utils@53.0.7.

Thanks to everyone who contributed to this release: