Fix wrong ordering when using unique in KNN.

[JulienPeloton]

We are using the wrong convention for scala.Ordering.get. From the doc we have

def gt(x: T, y: T): Boolean
  // Return true if x > y in the ordering.

while in the code we had:

// in BoundedUniquePriorityQueue.scala
if (head != null && ord.gt(a, head)) {
  // replace head by a
}

leading to wrong result!

without the fix:

With the fix:

[JulienPeloton]

Note that:

• I also rewrote takeOrdered with pattern matching instead of ifs.
• New unit tests have been added.

[JulienPeloton]

Note also that it was not seen without data repartitioning.

[codecov-io]

Codecov Report

:exclamation: No coverage uploaded for pull request base (queries@ad37750). Click here to learn what that means. The diff coverage is 90%.

@@            Coverage Diff             @@
##             queries     #111   +/-   ##
==========================================
  Coverage           ?   89.64%           
==========================================
  Files              ?       32           
  Lines              ?     1178           
  Branches           ?      194           
==========================================
  Hits               ?     1056           
  Misses             ?      122           
  Partials           ?        0

Flag	Coverage Δ
#python	91.24% <ø> (?)
#scala	88.95% <90%> (?)

Impacted Files	Coverage Δ
...com/spark3d/utils/BoundedUniquePriorityQueue.scala	90.47% <100%> (ø)
src/main/scala/com/spark3d/utils/Utils.scala	94.11% <88.88%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update ad37750...82bc7ae. Read the comment docs.