Towards improved FAIRness of the ThermoML Archive