terminal: catch the remaining stale ask/answer references

argus, reviewing PR#3793: a third stale ❓ ask mention survived in the
same file/table the first pass touched (docs/terminal-rendering.md's
icon legend) — tool_icon() has no ask/answer arm at all, confirmed by
reading the function directly. Swept the whole tree this time instead
of trusting the earlier narrow grep: found three more —
docs/web-ui/dashboard.md's S3TT1NGS section still documented the
expand-tool-output toggle as living on the dashboard, which moved to
the per-agent page's own SettingsMenu popover in #3780 and was never
followed up here; markdown.ts/streamRow.ts/terminal.css had the same
send/ask/answer/recv phrasing as the original two comments.

frontend/packages/agent/src/lib/markdown.ts

@@ -1,6 +1,6 @@
 // Markdown → sanitized HTML, ported from app.js's `mdNode`. Message
-// bodies rendered into the live stream (assistant text, send/ask/answer/
-// recv payloads) are untrusted (peer-agent / matrix-relayed content,
+// bodies rendered into the live stream (assistant text, send/recv
+// payloads) are untrusted (peer-agent / matrix-relayed content,
 // agent-authored files) — `marked` itself no longer sanitizes (v5+
 // dropped the built-in sanitizer), so every parse is run through
 // DOMPurify before it's ever handed to `dangerouslySetInnerHTML`.

frontend/packages/agent/src/lib/streamRow.ts

@@ -40,7 +40,7 @@ export interface StreamRow {
    * body today) — its own class, no markdown parsing. */
   childText?: StreamRowMeta;
   /** Sanitized-markdown body: appended under a flat row (assistant
-   * text) or inside an open details row (send/ask/answer/recv bodies). */
+   * text) or inside an open details row (send/recv bodies). */
   markdownBody?: string;
   /** Plain `<pre>` body inside a details row (generic long tool output). */
   plainBody?: string;