predict_julian_date_t is not a julian date

[bjorgan]

After working on #108: realized that our predict_julian_date_t is not strictly a Julian day count despite the name, since it is the number of days since December 31, 1979 rather than January 1, 4713 BC (Julian calendar). It is converted to a Julian day count proper in each function dealing with time by using the JULIAN_TIME_DIFF constant.

Should be unproblematic to modify the conversion function to return the value with the constant added, instead of leaving the adding until later. Would reduce some confusion and make the datatype naming more appropriate.