refactor(captain): route conversation completion by feature (#15317)

Conversation completion evaluations now use a dedicated internal LLM
feature with GPT 4.1 as the default. The internal route keeps the
completion model separate from the installation wide Captain model
override and from the assistant route, which can use GPT 5.2 for Captain
V2 accounts. Evaluations continue to use the installation API key and do
not consume Captain response credits.

## What changed

Added an internal `conversation_completion` feature to the LLM model
config and excluded internal features from account preferences, the
Captain settings API, and Super Admin model overrides.

Updated `Captain::ConversationCompletionService` to resolve its model
through `Llm::FeatureRouter`.

Added focused service and request coverage for model routing and
settings visibility.

enterprise/lib/captain/conversation_completion_service.rb

@@ -16,7 +16,8 @@ class Captain::ConversationCompletionService < Captain::BaseTaskService
     return default_incomplete_response('No messages found') if content.blank?
 
     response = make_api_call(
-      model: InstallationConfig.find_by(name: 'CAPTAIN_OPEN_AI_MODEL')&.value.presence || GPT_MODEL,
+      feature: 'conversation_completion',
+      model: self_hosted_model_override,
       messages: [
         { role: 'system', content: prompt_from_file('conversation_completion') },
         { role: 'user', content: content }
@@ -31,6 +32,12 @@ class Captain::ConversationCompletionService < Captain::BaseTaskService
 
   private
 
+  def self_hosted_model_override
+    return unless ChatwootApp.self_hosted_enterprise?
+
+    InstallationConfig.find_by(name: 'CAPTAIN_OPEN_AI_MODEL')&.value.presence
+  end
+
   def prompt_from_file(file_name)
     Rails.root.join('enterprise/lib/captain/prompts', "#{file_name}.liquid").read
   end