feat: support benchmarking vips processor

[knarewski]

Background

GdkPixbuf-based image processing can take over 1GB of memory for processing a single high-resolution photo. We've decided to introduce another processor, hoping that it'll improve Morandi's memory profile and performance.

Problems

Benchmarking scripts only supported pixbuf processor

Solution

• pass image processor id to Morandi.process in bin/process-single script - it adds support for vips processor and simplifies the code
• add vips processor to the benchmarking list in bin/benchmark-all script

Notes

I've noticed that vips outputs noticeably bigger (in terms of storage space) images, so I've also included output_size_mb in benchmarking results

Example benchmark

Observations

• memory used by vips is typically lower than by pixbuf; the difference gets very significant (eg 150MB vs 2.5GB RAM) for high-resolution images; that's expected, as pixbuf loads entire image to memory, while vips processes in chunks
• cpu consumption for vips is noticeably higher than pixbuf's; that can be controlled with concurrency level, but impacts rendering time; in order to make the time comparable between vips and pixbuf, vips concurrency has to be at least 2
• size of the output image for vips is roughly 10-50% higher than pixbuf's; I don't know the exact cause, I'd guess some implementation detail when saving; the size difference is apparent even on no-op read->write

Raw benchmark results

root@3a2291c480a4:/workspaces/morandi-rb# bundle exec bin/benchmark-full
Processing image: tmp/high-res-small-size-16000x11000px.jpg (Huge, pixelised greyscale gradient), 10 runs
Options: {"crop"=>"815,850,14909,10005", "straighten"=>0.5, "gamma"=>0.85}
  vips:
    real_time: avg 7.91; min 6.94; max 8.94
    kernel_time: avg 4.78; min 4.19; max 5.47
    user_time: avg 9.23; min 8.16; max 10.45
    cpu_percentage: avg 176; min 172; max 180
    rss_max_mb: avg 154.67; min 154.45; max 154.89
    output_size_mb: avg 4.12; min 4.12; max 4.12
  pixbuf:
    real_time: avg 7.37; min 7.02; max 8.08
    kernel_time: avg 1.26; min 1.14; max 1.41
    user_time: avg 5.93; min 5.66; max 6.47
    cpu_percentage: avg 97; min 97; max 97
    rss_max_mb: avg 2430.64; min 2430.45; max 2430.77
    output_size_mb: avg 2.47; min 2.47; max 2.47
Processing image: tmp/spider-8288816.jpg (10MB stock photo), 10 runs
Options: {"crop"=>"100,100,6000,4000", "angle"=>180, "straighten"=>-0.5, "gamma"=>1.2}
  vips:
    real_time: avg 2.96; min 2.73; max 3.45
    kernel_time: avg 0.33; min 0.29; max 0.42
    user_time: avg 3.37; min 3.15; max 4.04
    cpu_percentage: avg 124; min 122; max 129
    rss_max_mb: avg 178.26; min 177.95; max 178.59
    output_size_mb: avg 12.64; min 12.64; max 12.64
  pixbuf:
    real_time: avg 4.07; min 3.26; max 5.9
    kernel_time: avg 0.49; min 0.36; max 0.74
    user_time: avg 3.12; min 2.56; max 4.48
    cpu_percentage: avg 88; min 86; max 90
    rss_max_mb: avg 457.16; min 456.82; max 457.31
    output_size_mb: avg 10.03; min 10.03; max 10.03
Processing image: tmp/apple-8027938.jpg (1MB stock photo), 10 runs
Options: {"crop"=>"300,300,5000,3000", "gamma"=>1.0}
  vips:
    real_time: avg 2.31; min 1.92; max 2.9
    kernel_time: avg 0.21; min 0.15; max 0.27
    user_time: avg 2.0; min 1.71; max 2.51
    cpu_percentage: avg 95; min 94; max 97
    rss_max_mb: avg 155.61; min 155.44; max 155.82
    output_size_mb: avg 1.89; min 1.89; max 1.89
  pixbuf:
    real_time: avg 2.11; min 1.86; max 2.45
    kernel_time: avg 0.16; min 0.13; max 0.2
    user_time: avg 1.83; min 1.59; max 2.06
    cpu_percentage: avg 93; min 92; max 95
    rss_max_mb: avg 140.83; min 140.57; max 140.94
    output_size_mb: avg 1.5; min 1.5; max 1.5
Processing image: tmp/IMG_1425.jpg (A typical phone upload), 10 runs
Options: {"crop"=>"100,100,2500,2500", "straighten"=>0.5, "gamma"=>0.85}
  vips:
    real_time: avg 2.36; min 2.12; max 2.71
    kernel_time: avg 0.22; min 0.17; max 0.26
    user_time: avg 2.26; min 2.06; max 2.5
    cpu_percentage: avg 104; min 102; max 107
    rss_max_mb: avg 124.43; min 124.07; max 124.73
    output_size_mb: avg 1.59; min 1.59; max 1.59
  pixbuf:
    real_time: avg 2.28; min 2.07; max 2.69
    kernel_time: avg 0.21; min 0.16; max 0.27
    user_time: avg 1.96; min 1.77; max 2.32
    cpu_percentage: avg 94; min 94; max 96
    rss_max_mb: avg 243.4; min 243.2; max 243.57
    output_size_mb: avg 1.29; min 1.29; max 1.29