Improve CairoMakie thread safety

[jkrumbiegel]

cf #2391 and #812

After I added locks for font handles in JuliaGraphics/FreeTypeAbstraction.jl#87 I also add a few locks around font handling in CairoMakie. With that I don't get crashes anymore for this code:

function threadtest(n)
    fs = [scatter(1:10; axis = (; title = "$i")) for i in 1:n]
    @time Threads.@threads for i in 1:n
        save("temp/scatter_$i.png", fs[i])
    end
    return
end

threadtest(100)

For some reason I sometimes get hangs if I create the plots in threads as well, but that's not an issue from this PR which is only concerned with the rendering part. On an M1 macbook with 10 threads I get 3.109981 seconds (438.09 k allocations: 434.883 MiB, 0.85% gc time) for the example above without threads, and 0.437824 seconds (439.49 k allocations: 435.019 MiB) with threads enabled, which would be a speedup factor of 7.1.

[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	3.41s (3.37, 3.45) 0.02+-	169.25ms (166.07, 178.52) 4.37+-	476.51ms (472.86, 483.87) 3.84+-	7.50ms (7.27, 7.64) 0.14+-	26.03ms (25.93, 26.13) 0.08+-
master	3.41s (3.38, 3.46) 0.03+-	170.48ms (166.90, 176.70) 3.51+-	479.33ms (471.49, 495.26) 7.85+-	7.51ms (7.42, 7.58) 0.06+-	25.90ms (25.83, 25.96) 0.05+-
evaluation	1.00x invariant, -0.0s (-0.13d, 0.82p, 0.03std)	1.01x invariant, -1.23ms (-0.31d, 0.57p, 3.94std)	1.01x invariant, -2.82ms (-0.46d, 0.42p, 5.84std)	1.00x invariant, -0.01ms (-0.11d, 0.84p, 0.10std)	0.99x slower X, 0.13ms (2.13d, 0.00p, 0.06std)
CairoMakie	2.95s (2.92, 2.99) 0.03+-	163.82ms (161.35, 165.22) 1.33+-	100.60ms (98.17, 103.51) 1.65+-	7.22ms (7.10, 7.55) 0.16+-	889.78μs (885.45, 893.67) 3.10+-
master	2.95s (2.93, 2.99) 0.02+-	164.15ms (162.04, 165.99) 1.37+-	100.23ms (98.55, 102.89) 1.34+-	7.14ms (7.08, 7.21) 0.05+-	889.19μs (883.59, 892.58) 3.48+-
evaluation	1.00x invariant, 0.0s (0.05d, 0.93p, 0.03std)	1.00x invariant, -0.33ms (-0.24d, 0.66p, 1.35std)	1.00x invariant, 0.37ms (0.25d, 0.65p, 1.50std)	0.99x invariant, 0.08ms (0.68d, 0.24p, 0.10std)	1.00x invariant, 0.59μs (0.18d, 0.74p, 3.29std)
WGLMakie	3.57s (3.54, 3.65) 0.04+-	283.34ms (163.86, 371.50) 106.18+-	8.55s (8.42, 8.81) 0.14+-	9.34ms (9.07, 10.44) 0.49+-	68.80ms (67.56, 69.87) 0.94+-
master	3.55s (3.53, 3.58) 0.02+-	366.31ms (363.04, 373.83) 3.79+-	8.46s (8.41, 8.52) 0.05+-	9.15ms (9.12, 9.19) 0.03+-	68.33ms (66.92, 69.60) 0.82+-
evaluation	0.99x invariant, 0.02s (0.68d, 0.23p, 0.03std)	1.29x noisy🤷‍♀️, -82.97ms (-1.10d, 0.08p, 54.98std)	0.99x invariant, 0.09s (0.86d, 0.15p, 0.09std)	0.98x invariant, 0.19ms (0.54d, 0.35p, 0.26std)	0.99x invariant, 0.47ms (0.53d, 0.34p, 0.88std)