the best way to do this would be to weld on two metal bungs onto the tube, and get ceramic or teflon insulators that thread into that, and then install the probes into the insulators. a slightly less complicated way would be like that diagram you posted a few above. that is as "less complicated" as i would go with it though, i would personally never use epoxy in this way.
remember orientation of the tube will matter; you will always want the heater at the bottom so the sensor definately gets triggered before the element becomes unsubmerged. you also want to verify that the controller actually does register the signal change (in resistance- im guessing?) when it goes from being submerged to not being submerged. sometimes in a hot and humid environment, continuity sensors can detect a partial circuit even without being fully submerged, but the exact threshold can be incorrect and it may not register the fault.