refactor: move the Plex switch-token parser in with the Plex models

user_switch_response.dart was left holding a single 13-line function after
UserSwitchResponse's decorative fields were dropped, so the filename no
longer described its contents and it sat at lib/models/ root while every
other Plex model lives in lib/models/plex/.

Renamed to lib/models/plex/plex_switch_response.dart, with the test moved
alongside the other plex_*_test.dart files. The parser stays public so the
#1488 drift characterization tests keep exercising it directly.

lib/models/plex/plex_switch_response.dart

@@ -0,0 +1,13 @@
+/// INVARIANT (#1488): a successful token mint must never be lost to parsing
+/// of decorative fields. `authToken` is the only field any caller consumes
+/// (see plex_home_switch.dart), so nothing else on the `/switch` body is
+/// read. Plex has changed field shapes on this endpoint before (July 2026:
+/// profile language lists became CSV strings), and each drift used to brick
+/// token minting outright.
+String parsePlexSwitchAuthToken(Map<String, dynamic> json) {
+  final authToken = json['authToken'];
+  if (authToken is! String || authToken.isEmpty) {
+    throw const FormatException('Plex /switch response has no usable authToken');
+  }
+  return authToken;
+}