fix brack data_limits

[SimonDanisch]

#3179 introduced a regression in the data_limits for brackets, and since our reference images tests still have a way too high threshold, it didn't fail the tests.
@ffreyer, you think there are other places where this happened which are untested?
We should definitely look into the refimages this time before merging!

[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(display(fig))

	using	create	display	create	display
GLMakie	12.16s (12.09, 12.22) 0.05+-	1.20s (1.18, 1.23) 0.02+-	747.29ms (719.27, 801.82) 27.33+-	10.86ms (10.76, 11.03) 0.10+-	142.46ms (140.00, 144.12) 1.30+-
master	12.19s (12.06, 12.41) 0.12+-	1.20s (1.14, 1.24) 0.03+-	739.34ms (723.60, 752.85) 12.09+-	10.87ms (10.61, 11.04) 0.15+-	141.42ms (138.99, 143.88) 1.48+-
evaluation	-0.24%, -0.03s invariant (-0.32d, 0.57p, 0.08std)	+0.25%, 0.0s invariant (0.12d, 0.83p, 0.02std)	+1.06%, 7.96ms invariant (0.38d, 0.50p, 19.71std)	-0.13%, -0.01ms invariant (-0.11d, 0.84p, 0.13std)	+0.73%, 1.04ms invariant (0.74d, 0.19p, 1.39std)
CairoMakie	10.03s (9.81, 10.19) 0.14+-	1.21s (1.18, 1.25) 0.03+-	223.62ms (220.15, 231.15) 4.19+-	9.73ms (9.54, 10.00) 0.14+-	5.82ms (5.68, 6.02) 0.12+-
master	9.99s (9.75, 10.25) 0.16+-	1.16s (1.11, 1.21) 0.04+-	230.63ms (219.32, 242.00) 8.55+-	9.74ms (9.45, 10.14) 0.23+-	5.78ms (5.63, 5.98) 0.12+-
evaluation	+0.40%, 0.04s invariant (0.26d, 0.63p, 0.15std)	+4.15%, 0.05s slower X (1.45d, 0.02p, 0.03std)	-3.14%, -7.01ms invariant (-1.04d, 0.08p, 6.37std)	-0.07%, -0.01ms invariant (-0.03d, 0.95p, 0.19std)	+0.58%, 0.03ms invariant (0.28d, 0.61p, 0.12std)
WGLMakie	16.22s (16.07, 16.36) 0.11+-	1.66s (1.59, 1.78) 0.06+-	15.23s (15.02, 15.56) 0.20+-	21.98ms (21.12, 23.04) 0.78+-	1.40s (1.34, 1.45) 0.04+-
master	16.37s (16.08, 16.90) 0.27+-	1.70s (1.66, 1.74) 0.03+-	15.20s (15.01, 15.51) 0.17+-	22.13ms (20.67, 23.17) 0.85+-	1.40s (1.38, 1.43) 0.02+-
evaluation	-0.91%, -0.15s invariant (-0.71d, 0.22p, 0.19std)	-2.36%, -0.04s invariant (-0.81d, 0.17p, 0.04std)	+0.20%, 0.03s invariant (0.16d, 0.77p, 0.19std)	-0.70%, -0.15ms invariant (-0.19d, 0.73p, 0.81std)	+0.06%, 0.0s invariant (0.02d, 0.97p, 0.03std)

[ffreyer]

Makie.jl/test/makielayout.jl

Lines 381 to 405 in ac02141

	@testset "Axis scale" begin
	# This just shouldn't error
	try
	fig, ax, li = lines(1:10, 1:10)
	vlines!(ax, 3)
	hlines!(ax, 3)
	bp = barplot!(ax, 1 .+ 5 .* rand(10))
	vspan!(ax, 3, 4)
	hspan!(ax, 3, 4)
	bracket!(ax, 1, 1, 2, 2)
	eb = errorbars!(ax, 1:10, 1:10, [0.3 for _ in 1:10], whiskerwidth = 5)
	text!(ax, Point2f(2), text = "abba")
	tooltip!(ax, Point2f(8), "baab")
	tricontourf!(ax, 1 .+ 4 .* rand(5), 1 .+ 4 .* rand(5), rand(5))
	qqplot!(ax, 5:10, 1:5)
	ax.yscale = log10
	ax.yscale = identity
	ax.yscale = log10
	ax.yscale = identity
	@test true
	catch e
	@test false
	rethrow(e)
	end
	end

This one is erroring for me in #3040, possibly for the same reason. It includes most/all of the mixed-space plots and the limits look fine to me on this branch. So that's probably a good sign.
Maybe we can also extend this to check limits to catch these kind of issues in future? (This would probably need the update-on-display thing?)