Where does a reported client data model ride in MCP _meta?

**Documents:** `docs/public/guides/a2ui_over_mcp.md`,
`specification/v0_9_1/json/server_to_client.json` (v1.0:
`specification/v1_0/json/agent_to_renderer.json`)

A server asks for the client's data model with `sendDataModel` on `createSurface` or
`updateDataModel`. Over A2A the client answers with `a2uiClientDataModel` in transport metadata,
and the A2A extension names that placement.

We could not find the equivalent for MCP. The guide defines
`capabilities.a2ui.clientCapabilities` at initialize and `_meta.a2ui.clientCapabilities` per call,
and says nothing further about the reported data model. Did we miss it somewhere?

If it is not defined yet, which placement do you intend?

- `_meta["a2ui"]["clientDataModel"]`, gathering A2UI metadata under `a2ui` and dropping the prefix
from the inner names, the way the guide already does with `clientCapabilities`.
- `_meta["a2uiClientDataModel"]`, keeping the spec name at the root of `_meta`.

We had one implementation of each, and nothing surfaced. MCP `_meta` is a free form object, so the
reader that looked in the wrong place saw no key and no error.

Either choice works for us. Could the guide name one?

Read the linked guide and specification files to understand the A2UI over MCP protocol. Focus on the `_meta` object structure and how client capabilities are currently defined. Determine the intended placement for the reported client data model by comparing with the A2A extension's approach. Update the documentation to clarify the correct key path.

Escrito por el modelo de indexación a partir del texto del issue.