- compensation for rset != RSET_240K was wrong, it increased UVA power/ADC step for rset > RSET_240K
- UVA power/ADC step should have been decreasing as refresh time (sampling time) got longer
- same behavior as with integration_time setting, that works correctly
- this fixes a serious flaw where until now this library never powered the sensor down
- not noticable directly as the sensor still takes measurements but wastes power in between