Optimize vertical_sample performance

[Isotr0py]

• Similar to Optimize horizontal_sample performance #2894 with cache introduction
• But only enable pre-computed cache when memory budget is large enough, because the cache is inefficient in extreme case for vertical sample (low hit rate).

Benchmarks

resize 400x300 Nearest  time:   [2.1213 ms 2.1550 ms 2.1871 ms]
                        change: [-24.458% -22.829% -21.163%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 6 outliers among 100 measurements (6.00%)
  6 (6.00%) high mild

resize 400x300 Triangle time:   [6.9562 ms 7.0203 ms 7.0866 ms]
                        change: [-29.626% -28.783% -27.990%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 9 outliers among 100 measurements (9.00%)
  9 (9.00%) high mild

resize 400x300 CatmullRom
                        time:   [9.7698 ms 9.8562 ms 9.9452 ms]
                        change: [-43.885% -43.137% -42.386%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 2 outliers among 100 measurements (2.00%)
  2 (2.00%) high mild

resize 400x300 Gaussian time:   [12.528 ms 12.639 ms 12.754 ms]
                        change: [-49.790% -49.195% -48.586%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 1 outliers among 100 measurements (1.00%)
  1 (1.00%) high mild

resize 400x300 Lanczos3 time:   [12.608 ms 12.724 ms 12.843 ms]
                        change: [-49.570% -48.980% -48.370%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 2 outliers among 100 measurements (2.00%)
  2 (2.00%) high mild

large/resize 2000x2000 Nearest
                        time:   [63.963 ms 64.532 ms 64.927 ms]
                        change: [-10.545% -9.1141% -7.7965%] (p = 0.00 < 0.05)
                        Performance has improved.
large/resize 2000x2000 Triangle
                        time:   [72.503 ms 72.999 ms 73.669 ms]
                        change: [-16.895% -15.550% -14.281%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 1 outliers among 10 measurements (10.00%)
  1 (10.00%) low mild
large/resize 2000x2000 CatmullRom
                        time:   [80.833 ms 81.329 ms 82.067 ms]
                        change: [-25.137% -24.118% -23.064%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 10 measurements (30.00%)
  1 (10.00%) low mild
  2 (20.00%) high mild
Benchmarking large/resize 2000x2000 Gaussian: Warming up for 1.0000 s
Warning: Unable to complete 10 samples in 5.0s. You may wish to increase target time to 5.0s or enable flat sampling.
large/resize 2000x2000 Gaussian
                        time:   [87.334 ms 88.567 ms 89.422 ms]
                        change: [-29.560% -28.375% -27.306%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 1 outliers among 10 measurements (10.00%)
  1 (10.00%) high mild
Benchmarking large/resize 2000x2000 Lanczos3: Warming up for 1.0000 s
Warning: Unable to complete 10 samples in 5.0s. You may wish to increase target time to 5.1s or enable flat sampling.
large/resize 2000x2000 Lanczos3
                        time:   [91.461 ms 92.507 ms 93.475 ms]
                        change: [-26.768% -25.409% -23.994%] (p = 0.00 < 0.05)
                        Performance has improved.