Examples work again.
- removed dependency on Guava (which it didn't actually depend on) and on Piggybank (since date functions are now first class)
- path to datafu jar correct for current repo layout
- The quantile examples demonstrate a comparison of the exact vs approx algorithms (my first question was "how approximate is approximate?" 0.5% to 1.5% on the little drip of data that's using)
- added the script to generate data for quantile and a frozen version of 30k records.
- Quantile examples demonstrate both ways of constructing a Quantile UDF (number of partitions vs list of breakpoints)
There are two patches attached. One adds a 30k-line data file. The other holds the changes to the code and so forth.