Alternative approach for defining named UDFs.

[piotrszul]

trafficstars

See branch: spike/340 Package: au.csiro.pathling.sql.udfs;

This approach uses annotations rather then Spring beans and my be more appropriate if these the UDFS (such as datetime or quantity operations) needed to be exposed from python interface.

Also UDFs should be more focused on unit testing the UDF operations themselves rather that the FhirPaths that use them. Or at least it would be probably easier to write more through tests cases for the UDFs (regardless if they are beans or functions) than for FhirPaths. And also I assume that in the future we my want/need to expose these UDFs in the python API (so perhaps they should be tested independently from FhirPaths).