Better handle hash collisions by using Java LinkedHashMap

[gmethvin]

Switch the underlying map implementation to use a Java LinkedHashMap with the Scala wrapper, to take advantage of the Java implementation's more efficient handling of hash collisions where the keys are Comparable.

Fixes #186.

[gmethvin]

Parsing an object with 10000 hash collisions (-i 10 -wi 5 -f1 -t4):

before:

[info] # Warmup Iteration   1: 6.028 ops/s
[info] # Warmup Iteration   2: 7.308 ops/s
[info] # Warmup Iteration   3: 7.593 ops/s
[info] # Warmup Iteration   4: 7.768 ops/s
[info] # Warmup Iteration   5: 7.949 ops/s
[info] Iteration   1: 6.925 ops/s
[info] Iteration   2: 7.489 ops/s
[info] Iteration   3: 7.810 ops/s
[info] Iteration   4: 7.867 ops/s
[info] Iteration   5: 7.485 ops/s
[info] Iteration   6: 7.732 ops/s
[info] Iteration   7: 7.634 ops/s
[info] Iteration   8: 7.656 ops/s
[info] Iteration   9: 7.571 ops/s
[info] Iteration  10: 7.343 ops/s

after:

[info] # Warmup Iteration   1: 297.834 ops/s
[info] # Warmup Iteration   2: 538.602 ops/s
[info] # Warmup Iteration   3: 660.174 ops/s
[info] # Warmup Iteration   4: 698.398 ops/s
[info] # Warmup Iteration   5: 783.469 ops/s
[info] Iteration   1: 761.133 ops/s
[info] Iteration   2: 788.541 ops/s
[info] Iteration   3: 789.512 ops/s
[info] Iteration   4: 775.890 ops/s
[info] Iteration   5: 778.181 ops/s
[info] Iteration   6: 767.591 ops/s
[info] Iteration   7: 726.882 ops/s
[info] Iteration   8: 764.094 ops/s
[info] Iteration   9: 764.370 ops/s
[info] Iteration  10: 769.734 ops/s

[marcospereira]

Thanks, @gmethvin.

[gmethvin]

Does someone want to backport and make a new 2.6 release?