Improve performance for small inputs

[gkowal]

By reusing the buffer the small input performance improves dramatically.

The example output on my machine:

BenchmarkFastCDCSize/1k                  5203628               228 ns/op        4492.41 MB/s             1.00 chunks           1 B/op          0 allocs/op
BenchmarkFastCDCSize/4k                  1750215               686 ns/op        5974.14 MB/s             1.00 chunks           4 B/op          0 allocs/op
BenchmarkFastCDCSize/16k                  455090              2488 ns/op        6584.83 MB/s             1.00 chunks          18 B/op          0 allocs/op
BenchmarkFastCDCSize/32k                  238017              4860 ns/op        6741.91 MB/s             1.00 chunks          35 B/op          0 allocs/op
BenchmarkFastCDCSize/64k                  122582              9601 ns/op        6825.77 MB/s             1.00 chunks          68 B/op          0 allocs/op
BenchmarkFastCDCSize/128k                  61484             19401 ns/op        6755.81 MB/s             1.00 chunks         136 B/op          0 allocs/op
BenchmarkFastCDCSize/256k                  30966             38856 ns/op        6746.52 MB/s             1.00 chunks         270 B/op          0 allocs/op
BenchmarkFastCDCSize/512k                   2385            483351 ns/op        1084.69 MB/s             1.00 chunks        3517 B/op          0 allocs/op
BenchmarkFastCDCSize/1M                      786           1493472 ns/op         702.11 MB/s             1.00 chunks       10672 B/op          0 allocs/op
BenchmarkFastCDCSize/4M                      172           6929246 ns/op         605.30 MB/s             4.00 chunks       48771 B/op          0 allocs/op
BenchmarkFastCDCSize/16M                      43          25280504 ns/op         663.64 MB/s            15.0 chunks       195086 B/op          0 allocs/op
BenchmarkFastCDCSize/32M                      21          51466408 ns/op         651.97 MB/s            28.0 chunks       399462 B/op          0 allocs/op
BenchmarkFastCDCSize/64M                      10         104412456 ns/op         642.73 MB/s            56.0 chunks       838872 B/op          0 allocs/op
BenchmarkFastCDCSize/128M                      5         207613783 ns/op         646.48 MB/s           116 chunks        1677744 B/op          0 allocs/op
BenchmarkFastCDCSize/512M                      2         839211188 ns/op         639.73 MB/s           454 chunks        4194360 B/op          1 allocs/op
BenchmarkFastCDCSize/1G                        1        1727176911 ns/op         621.67 MB/s           902 chunks        8388720 B/op          2 allocs/op

This PR includes also PR #1, therefore the previous one can be ignored.

[codecov-commenter]

Codecov Report

Merging #2 into master will increase coverage by 0.28%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master       #2      +/-   ##
==========================================
+ Coverage   95.83%   96.11%   +0.28%     
==========================================
  Files           1        1              
  Lines          96      103       +7     
==========================================
+ Hits           92       99       +7     
  Misses          2        2              
  Partials        2        2              

Impacted Files	Coverage Δ
fastcdc.go	96.11% <100.00%> (+0.28%)	⬆️

---

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 f27242f...3cac44c. Read the comment docs.