benchmark: add --analyze option to benchmark/scatter.js

[jasnell]

Similar to compare.js, provide an R-free --analyze to make scatter.js more self-contained.

Allows scatter.js to produce data without requiring the R script to be used.

$ node benchmark/scatter.js --analyze --xaxis chunkLen --category encoding benchmark/string_decoder/string-decoder.js

chunkLen  encoding   samples          rate  confidence.interval        median    median.interval
      16  'ascii'          6   2,248,111.0    61,570.3 (±2.74%)   2,241,996.7   [-3.75%, +4.47%]
      16  'utf16le'        6   2,004,145.9    33,363.7 (±1.66%)   1,990,015.9   [-1.03%, +3.44%]
      16  'utf8'           6   1,156,990.9    84,288.2 (±7.29%)   1,186,284.7  [-16.23%, +1.15%]  (!)
     256  'ascii'          6  11,271,912.4   398,786.9 (±3.54%)  10,991,001.5   [-0.47%, +7.61%]
     256  'utf16le'        6  10,243,336.8   792,495.2 (±7.74%)   9,908,838.3  [-4.73%, +17.25%]  (!)
     256  'utf8'           6   8,247,482.7   316,510.5 (±3.84%)   8,282,009.5   [-5.80%, +3.19%]
    1024  'ascii'          6  11,130,650.9   438,019.6 (±3.94%)  10,838,937.5   [-1.34%, +7.10%]
    1024  'utf16le'        6   9,786,274.4   186,207.1 (±1.90%)   9,690,675.1   [-0.77%, +4.20%]
    1024  'utf8'           6   8,479,690.1   405,762.8 (±4.79%)   8,525,516.7   [-9.01%, +4.06%]  (!)

Rate in operations/second; longer is faster. │ marks the mean and the
shaded band (░) is its 95% confidence interval, so bars whose bands
overlap are not clearly different.

                                  0                           11,670,699.3
                                  +--------------------------------------+
chunkLen=16 encoding='ascii'      ███████│                                   2,248,111.0
chunkLen=16 encoding='utf16le'    ██████│                                    2,004,145.9
chunkLen=16 encoding='utf8'       ███│                                       1,156,990.9

chunkLen=256 encoding='ascii'     █████████████████████████████████████░│░  11,271,912.4
chunkLen=256 encoding='utf16le'   ████████████████████████████████░░░│░░    10,243,336.8
chunkLen=256 encoding='utf8'      ███████████████████████████░│              8,247,482.7

chunkLen=1024 encoding='ascii'    █████████████████████████████████████░│░  11,130,650.9
chunkLen=1024 encoding='utf16le'  █████████████████████████████████│         9,786,274.4
chunkLen=1024 encoding='utf8'     ████████████████████████████░│             8,479,690.1

Change between consecutive chunkLen values (Mann-Whitney U, Cliff's delta):

  encoding='ascii'
    16 -> 256    +401.39%  16.0x  exponent=+0.58  p=0.0039  delta=+1.000 (large)
    256 -> 1024    -1.25%   4.0x  exponent=-0.01  p=0.3367  delta=-0.333 (medium)

  encoding='utf16le'
    16 -> 256    +411.11%  16.0x  exponent=+0.59  p=0.0039  delta=+1.000 (large)
    256 -> 1024    -4.46%   4.0x  exponent=-0.03  p=0.1495  delta=-0.500 (large)

  encoding='utf8'
    16 -> 256    +612.84%  16.0x  exponent=+0.71  p=0.0039  delta=+1.000 (large)
    256 -> 1024    +2.82%   4.0x  exponent=+0.02  p=0.2002  delta=+0.444 (medium)

[nodejs-github-bot]

Review requested:

• @nodejs/performance
• @nodejs/tsc

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/76594/

[mcollina]

RSLGTM