chore: restructure llm plugin to be more coherent & fix various bugs in llm ui

Plugins/ZShell/Llm/messagemodel.cpp

@@ -40,14 +40,18 @@ int ChatMessageModel::rowOf(const ChatMessage* message) const {
 
 ChatMessage* ChatMessageModel::createMessage(
 	ChatMessage::Role role, qint64 timestamp) {
-	return new ChatMessage(role, timestamp, this);
+	return new ChatMessage(m_session, role, timestamp, this);
 }
 
 ChatMessage* ChatMessageModel::appendNewest(
 	ChatMessage::Role role, const QString& content, qint64 timestamp) {
-	auto* message = new ChatMessage(role, timestamp, this);
-	auto* generation = message->addGeneration(timestamp);
-	generation->setContent(content);
+	auto* message = createMessage(role, timestamp);
+	if (role == ChatMessage::Role::User) {
+		message->setContent(content);
+	} else {
+		auto* generation = message->addGeneration(timestamp);
+		generation->setContent(content);
+	}
 
 	beginInsertRows(QModelIndex(), 0, 0);
 	m_messages.prepend(message);