Zen master-of-zen

Goals and motivation

Current fast scene detection in rav1e relatively slow to compared methods of scene detection, and have proclivity to show false results or don't detect scene changes where they appear

This pr reworks fast scene detection algorithm, making it faster, better, and more accurate

Achieved goals are:

Faster decision making ( Both less and more efficient computations )
More accurate scene detection, by adjusting threshold based on previous frames
Frame downscale for faster decisions.

  for x in 1..=scale {
    let src_row = &data_origin[(src.cfg.stride * 2 + x)..];
    let local_sum: usize = (1..scale)
      .into_par_iter()
      .map(|y| u32::cast_from(src_row[col * 2 + y]) as usize)
      .sum();
    sum += local_sum;
  }

let mut sum = 0;

for x in 1..=scale {
  let src_row = &data_origin[(src.cfg.stride * 2 + x)..];
  sum += (1..scale)
    .into_par_iter()
    .for_each(|y| u32::cast_from(src_row[col * 2 + y]) as usize)
    .sum();
 }

let mut sum = 0;

for x in 1..=scale {
  let src_row = &data_origin[(src.cfg.stride * 2 + x)..];
  for y in 1..=scale {
    sum += u32::cast_from(src_row[col * 2 + y]) as usize
  }
}

4k

720p

Main changes

Y plane downscaling before comparison
Frame buffer for keeping scaled frames as long as they required

2160p

Old

rav1e --rdo-lookahead-frames research

Motivation:

Make speed settings faster with reasonable trade-off on different rdo-lookahead-frames values and suggest new defaults for each speed

	# Fields meanings: <source root>/av1/encoder/firstpass.h
	fields = [
	"frame",
	"weight",
	"intra_error",
	"frame_avg_wavelet_energy",
	"coded_error",
	"sr_coded_error",
	"tr_coded_error",
	"pcnt_inter",

	#!/usr/bin/env python

	from matplotlib import pyplot as plt
	import matplotlib

	matplotlib.rc('xtick', labelsize=24)
	matplotlib.rc('ytick', labelsize=24)

	data = {
	10: {

	#!/usr/bin/env python

	from matplotlib import pyplot as plt
	import matplotlib

	matplotlib.rc('xtick', labelsize=24)
	matplotlib.rc('ytick', labelsize=24)

	data = {
	10: {