search_dom returns attributes but not text content #516

Summary

search_dom's output schema is {tag, nodeId, backendNodeId, attributes} (src/browser/dom.ts:79-87) — no text content. An agent can locate an element (e.g. a <script type="application/json" id="..."> by id/selector) but then has no way to read what's inside it.

Current workarounds and their cost

• get_dom: writes the entire document's outerHTML to a temp file (src/browser/dom.ts:40-49) — works, but is a whole-document read for a single element's text.
• evaluate_script: works, but requires the agent to write and run arbitrary JS just to read .textContent.

Proposal

An opt-in includeText (or similar) parameter on search_dom that adds each matched node's textContent to the result. Closes the loop cheaply for the common case of "I found the right element, now let me read it" without a full-document dump.