Write colorbuffer directly to io if it fits in format

[jkrumbiegel]

Noticed with @profview that record seemed to spend quite a bit of time in a copy! call. As this might not be necessary under many circumstances, I've added a branch that avoids copy!, leading to a nice speedup of about 20% in this simple example:

using GLMakie

# before
julia> @time record(lines(cumsum(randn(1000))), "test.mp4", 1:1000, framerate = 60) do i
       end
  6.611545 seconds (1.05 M allocations: 38.077 MiB, 0.24% gc time, 0.28% compilation time)

# after
julia> @time record(lines(cumsum(randn(1000))), "test.mp4", 1:1000, framerate = 60) do i
       end
  5.553962 seconds (1.05 M allocations: 37.878 MiB, 0.33% compilation time)
"test.mp4"

[MakieBot]

Compile Times benchmark

Note, that these numbers may fluctuate on the CI servers, so take them with a grain of salt. All benchmark results are based on the mean time and negative percent mean faster than the base branch. Note, that GLMakie + WGLMakie run on an emulated GPU, so the runtime benchmark is much slower. Results are from running:

using_time = @ctime using Backend
# Compile time
create_time = @ctime fig = scatter(1:4; color=1:4, colormap=:turbo, markersize=20, visible=true)
display_time = @ctime Makie.colorbuffer(display(fig))
# Runtime
create_time = @benchmark fig = scatter(1:4; color=1:4, colormap=:turbo, markersize=20, visible=true)
display_time = @benchmark Makie.colorbuffer(fig)

	using	create	display	create	display
GLMakie	5.03s (4.97, 5.07) 0.04+-	108.72ms (107.54, 110.13) 0.84+-	580.91ms (409.70, 657.38) 116.58+-	8.48ms (8.45, 8.51) 0.02+-	25.71ms (25.63, 25.87) 0.09+-
master	5.07s (5.00, 5.16) 0.05+-	108.16ms (107.51, 108.76) 0.54+-	613.62ms (410.02, 653.70) 89.84+-	8.28ms (8.20, 8.37) 0.05+-	25.71ms (25.58, 26.12) 0.20+-
evaluation	1.01x invariant, -0.03s (-0.76d, 0.19p, 0.04std)	0.99x invariant, 0.56ms (0.78d, 0.17p, 0.69std)	1.06x noisy🤷‍♀️, -32.71ms (-0.31d, 0.57p, 103.21std)	0.98x slower X, 0.2ms (4.81d, 0.00p, 0.04std)	1.00x invariant, -0.0ms (-0.01d, 0.99p, 0.14std)
CairoMakie	4.87s (4.83, 4.91) 0.03+-	109.92ms (109.35, 110.99) 0.54+-	166.96ms (166.38, 167.51) 0.38+-	9.40ms (9.13, 9.60) 0.22+-	1.22ms (1.17, 1.24) 0.02+-
master	4.84s (4.80, 4.89) 0.03+-	113.14ms (109.29, 118.40) 4.45+-	167.43ms (167.11, 168.07) 0.32+-	9.21ms (8.98, 9.51) 0.21+-	1.19ms (1.14, 1.24) 0.04+-
evaluation	0.99x slower X, 0.04s (1.23d, 0.04p, 0.03std)	1.03x invariant, -3.22ms (-1.02d, 0.10p, 2.50std)	1.00x faster ✓, -0.47ms (-1.33d, 0.03p, 0.35std)	0.98x invariant, 0.19ms (0.90d, 0.12p, 0.21std)	0.97x invariant, 0.03ms (1.01d, 0.09p, 0.03std)
WGLMakie	5.63s (5.52, 5.81) 0.13+-	113.38ms (110.10, 124.01) 4.95+-	4.76s (4.62, 5.04) 0.16+-	12.13ms (11.53, 14.11) 0.98+-	118.20ms (116.32, 120.08) 1.26+-
master	5.63s (5.53, 5.81) 0.12+-	112.84ms (109.51, 117.72) 3.58+-	5.08s (4.90, 5.45) 0.19+-	12.01ms (11.61, 12.73) 0.45+-	120.23ms (118.70, 123.06) 1.49+-
evaluation	1.00x invariant, -0.0s (-0.02d, 0.97p, 0.12std)	1.00x invariant, 0.54ms (0.13d, 0.82p, 4.27std)	1.07x faster✅, -0.32s (-1.80d, 0.01p, 0.18std)	0.99x invariant, 0.12ms (0.16d, 0.77p, 0.72std)	1.02x faster ✓, -2.02ms (-1.47d, 0.02p, 1.38std)