String.split/1 refactor - #15781

prop_path = "#{unicode_folder}/PropList.txt"
prop = File.read!(prop_path)
lines = String.split(prop, "\n")
words = String.split(prop) |> Enum.chunk_every(2) |> Enum.map(&Enum.join(&1," "))

Benchee.run(
  %{
    "binary_split" => fn input -> Enum.map(input, &String.Break.split/1) end,
    "offset - proposed" => fn input -> Enum.map(input, &String.BreakNew.split/1) end,

  },
  inputs: [
    {"words", words},
    {"lines", lines},
    {"file", [prop]}
  ],
  time: 1,
  memory_time: 1
) 

##### With input words #####
Name                        ips        average  deviation         median         99th %
offset - proposed        1.02 K        0.98 ms    ±27.56%        0.93 ms        1.70 ms
binary_split           0.0355 K       28.18 ms     ±4.21%       28.56 ms       30.86 ms

Comparison: 
offset - proposed        1.02 K
binary_split           0.0355 K - 28.69x slower +27.20 ms

Memory usage statistics:

Name                 Memory usage
offset - proposed       812.03 KB
binary_split            927.26 KB - 1.14x memory usage +115.23 KB

**All measurements for memory usage were the same**

##### With input lines #####
Name                        ips        average  deviation         median         99th %
offset - proposed        1.39 K        0.72 ms    ±14.94%        0.76 ms        0.94 ms
binary_split            0.162 K        6.18 ms     ±5.65%        6.14 ms        7.66 ms

Comparison: 
offset - proposed        1.39 K
binary_split            0.162 K - 8.57x slower +5.46 ms

Memory usage statistics:

Name                 Memory usage
offset - proposed       402.42 KB
binary_split            788.15 KB - 1.96x memory usage +385.73 KB

**All measurements for memory usage were the same**

##### With input file #####
Name                        ips        average  deviation         median         99th %
binary_split             1.20 K      830.18 μs    ±30.29%      738.89 μs     1748.19 μs
offset - proposed        1.11 K      898.37 μs    ±17.23%      975.08 μs     1229.29 μs

Comparison: 
binary_split             1.20 K
offset - proposed        1.11 K - 1.08x slower +68.19 μs

Memory usage statistics:

Name                 Memory usage
binary_split              1.27 KB
offset - proposed       296.11 KB - 232.53x memory usage +294.84 KB

**All measurements for memory usage were the same**