Adjust layout of ReferenceUpdater

[ffreyer]

Description

This changes the layout of the page generated by ReferenceUpdater to be a table. The format is:

	GLMakie	CairoMakie	WGLMakie
test 1	image	image	image
test 2	image	image	image

without having the column and row labels explicitly. If a refimg does not exist for a given backend, the cell is left blank. These change are meant to make it easier to compare backend to backend and also bring cross-backend failures closer together. (E.g. if something is changed in Makie requiring refimg X to be updated for all backends, the related failures are now grouped together.)

Refimgs now resize based on the table column width (up to 100% of their size). This is sort of a consequence of a layout, but I think it's also generally makes refimages easier to check.

If a refimg causes CI to fail (score > 0.05) the background becomes red, otherwise yellow (> 0.001) or light gray (else). This should make it easier to spot which test in a row is failing and it also differentiates the page/cell background from the figure background.

Sorting is now based on maximum(row_scores) to move all the worst scores to the top of the page.

Type of change

• dev tools

[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	4.43s (4.35, 4.57) 0.08+-	106.15ms (104.14, 113.31) 3.23+-	472.35ms (464.98, 488.48) 8.10+-	9.41ms (9.24, 9.72) 0.15+-	26.08ms (25.70, 26.73) 0.44+-
master	4.37s (4.33, 4.41) 0.03+-	105.09ms (103.44, 107.84) 1.56+-	467.74ms (463.78, 473.19) 3.07+-	9.34ms (8.93, 9.51) 0.19+-	25.91ms (25.69, 26.12) 0.15+-
evaluation	0.99x invariant, 0.06s (1.01d, 0.10p, 0.05std)	0.99x invariant, 1.07ms (0.42d, 0.45p, 2.40std)	0.99x invariant, 4.61ms (0.75d, 0.20p, 5.59std)	0.99x invariant, 0.08ms (0.44d, 0.43p, 0.17std)	0.99x invariant, 0.17ms (0.51d, 0.37p, 0.30std)
CairoMakie	4.46s (4.40, 4.51) 0.04+-	121.62ms (117.16, 128.61) 3.66+-	186.74ms (181.57, 195.88) 5.51+-	10.52ms (9.85, 10.72) 0.31+-	1.20ms (1.19, 1.21) 0.01+-
master	4.49s (4.42, 4.53) 0.05+-	122.91ms (117.25, 129.81) 5.04+-	192.45ms (184.27, 201.35) 6.82+-	10.44ms (10.12, 10.87) 0.26+-	1.24ms (1.21, 1.26) 0.02+-
evaluation	1.01x invariant, -0.03s (-0.71d, 0.21p, 0.04std)	1.01x invariant, -1.3ms (-0.29d, 0.59p, 4.35std)	1.03x invariant, -5.71ms (-0.92d, 0.11p, 6.17std)	0.99x invariant, 0.08ms (0.28d, 0.60p, 0.28std)	1.03x faster ✓, -0.04ms (-2.46d, 0.00p, 0.01std)
WGLMakie	4.83s (4.70, 4.99) 0.09+-	111.38ms (105.86, 115.82) 3.68+-	9.31s (8.99, 9.67) 0.26+-	12.69ms (11.53, 17.87) 2.30+-	118.31ms (113.67, 123.28) 3.06+-
master	4.81s (4.73, 4.95) 0.09+-	111.67ms (106.19, 121.08) 5.43+-	9.71s (9.46, 10.12) 0.23+-	12.05ms (11.68, 12.97) 0.46+-	118.05ms (115.43, 119.43) 1.50+-
evaluation	1.00x invariant, 0.02s (0.21d, 0.70p, 0.09std)	1.00x invariant, -0.29ms (-0.06d, 0.91p, 4.55std)	1.04x faster ✓, -0.41s (-1.67d, 0.01p, 0.24std)	0.95x noisy🤷‍♀️, 0.64ms (0.38d, 0.50p, 1.38std)	1.00x invariant, 0.27ms (0.11d, 0.84p, 2.28std)