docs: fix get_metadata page-enumeration guidance (returns only the active page)

[jkRaccoon]

What

get_metadata with no nodeId is documented — both in this figma-use gotcha and in the tool's own description — as returning the document's full list of top-level pages. In practice it returns only the currently-active page.

Reproduction (2026-07-10)

Called get_metadata with only a fileKey (no nodeId) against a file that has 9 pages. Full response:

No nodeId was provided. Listing the top-level pages of the document. Call get_metadata again with one of the page ids below (or any node id underneath) to get the XML metadata for that subtree.

Top-level pages of the document:
- 0:1: COVER

Only the active page (COVER) came back; the other 8 pages (login, project, estimate, …) were omitted. figma.root.children via use_figma returns all 9.

Why it matters

The tool description says:

The nodeId parameter is optional: when omitted, the tool returns a list of the top-level pages (guid + name) in the document

An agent that trusts this makes a call expecting all pages, receives one, and has to fall back to use_figma + figma.root.children anyway — a wasted tool call (and tokens) on every session that needs full-document page enumeration.

This PR

Rewrites the gotcha to describe the actual behavior and points to use_figma + figma.root.children as the only reliable enumeration path.

-- Call `get_metadata` with **no nodeId** — it returns the document's top-level pages as `{guid, name}` entries (no XML dump). This is the cheapest way to discover pages.
-- For more detail per page (e.g. child counts, top-level node types), fall back to `use_figma`:
+- **Do not rely on `get_metadata` to enumerate pages.** Even with **no nodeId** ... it returns only the **currently-active page**, not all of them ...
+- **Enumerate pages with `use_figma` — this is the only reliable way:**

Also — server-side fix requested

This PR can only fix the guide doc. The get_metadata tool description itself (server-side, not in this repo) carries the same incorrect claim, and the response header ("Listing the top-level pages of the document") is likewise misleading. Please also either:

1. make get_metadata with no nodeId actually return all top-level pages, or
2. correct the description and the response header to say it returns only the active page and point to use_figma for full enumeration.

Filing this note here because Issues are disabled on this repo — happy to move it wherever the team prefers.