Slate v2 richtext renderSegment should read text-node marks from the snapshot

What happened

EditableText already knew how to split one text node into rendered segments from projection slices.

That was enough for decoration-style rendering.

It was not enough for current richtext rendering, because renderSegment(...) had no access to the text node's mark props like bold, italic, or code.

What fixed it

The rendered segment now carries the current text node's non-text props as marks.

That data is read from the snapshot by path or runtime id before the segments are rendered.

Why this works

Marks belong to the text node, not to the projection system.

Projection slices tell the renderer where to split. The snapshot text node tells the renderer how that leaf should be formatted.

Using both gives one clean render seam:

• slices decide segment boundaries
• text-node props decide mark wrappers

Reusable rule

For current slate-react-v2 richtext rendering:

• use projection slices to split leaves
• use snapshot text-node props to expose leaf marks
• do not pretend the projection layer alone can recover text formatting