Fix detector_tensor caching

This is an issue that came up when I was trying to code up an example for our Python group. Setting latitude, longitude, x/yarm_azimuth, and x/yarm_tilt did not update the the detector_tensor if x and y weren't accessed in between. The change makes sure that x and y are up to date when detector_tensor is evaluated and any change in the aforementioned attributes will trigger an update in detector_response rather than returning the cached value.