feat(tts): add websocket transport option for xAI TTS (#548)

* feat(tts): add websocket transport option for xAI TTS

xAI's pipecat service ships two implementations: XAIHttpTTSService
(batch REST, used by the existing xAI integration) and XAITTSService
(realtime WebSocket streaming). This adds a transport field on
XAITTSConfiguration ("http", default, unchanged behavior | "websocket")
and wires the latter into create_tts_service(), so live voice calls can
opt into lower time-to-first-byte streaming instead of request/response.

Originally built and battle-tested independently against v1.41.0 in a
production deployment (7 live Telnyx calls, 2026-07-15) before this PR
existed; ported onto current main and given a config toggle instead of
a second provider entry so it composes with the existing xAI HTTP path
rather than duplicating it.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

* refactor(tts): use websocket-only xAI TTS, drop the transport toggle

---------

Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com>
Co-authored-by: Sabiha Khan <sabihak89@gmail.com>

ui/package-lock.json

@@ -1,12 +1,12 @@
 {
   "name": "ui",
-  "version": "1.41.0",
+  "version": "1.42.0",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "ui",
-      "version": "1.41.0",
+      "version": "1.42.0",
       "dependencies": {
         "@calcom/embed-react": "^1.5.3",
         "@dagrejs/dagre": "^1.1.4",