I used an easy workaround. Leaving it here for future.
- Change the name of the odom_frame and base_link_frame parameter used by RosAria to publish tf from the default odom/base_link to say Aria_odom/Aria_base_link for example.
- Next, write a small rospy snippet to subscribe to /RosaAria/pose messages and rename the frame_id back to whatever frames you are using for robot_localization. Publish this new message from a new topic and use this topic for odom_n in robot_localization config file. Run this snippet before launching robot_localization.
The above ensures that there is no clashing transform in the the frames used by robot_localization while still allowing the odom data to be used.