Update benchmark results (#94)

Author: catap

CHANGELOG.md

@@ -3,6 +3,7 @@
 All notable changes to this project will be documented in this file.
 
 ## [unreleased]
+- Performance improved: now it's about 30% slower than C version
 
 ## [3.1.0] - 2022-05-17
 - Introduced `doneCallback` and `doneXorCallBack`

README.md

@@ -71,7 +71,7 @@ All benchmarks were performed on two machines:
 Short summary:
  - it is about 4 times slower than [AVX2 assembly] version via JNI which is
    expected,
- - it is about 30% slower than the original C version via JNI,
+ - it is about 20% slower than the original C version via JNI,
  - it has a constant memory footprint (yeah, no GC on hashing!),
  - increasing result hash size has the same impact as hashing.