display fixes

[SimonDanisch]

So, GLMakie not opening a window with activated Plotpane really was not a great idea...
Lots of complaints, and I also tripped over this quite a bit.
This PR changes this back to opening a window by default for GLMakie!
But this means, that one now needs to call GLMakie.activate!(inline=true) to get GLMakie back into the plotpane.

This PR also changes a few other things:

• can_show_inline(backend) was weirdly buggy with VSCode, a.k.a most of the time it worked that e.g. WGLMakie would show in the plotpane if the plotpane is enabled, but sometimes it just stopped working...Not sure why, but looking at the VSCode code, it should never work the way can_show_inline is implemented (not checking for MIME"juliavscode/html") - but it definitely worked correctly most of the time 🤷
• now one can do display(plot; inline=false) to force a window even if inline is true. Of course this doesn't work the other way around, since VSCode somehow doesn't get into the show path anymore with calling display(fig; inline=true) directly... Not sure why...

[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	10.73s (10.65, 10.79) 0.06+-	1.04s (1.03, 1.06) 0.01+-	769.22ms (756.80, 821.44) 23.52+-	9.88ms (9.79, 9.94) 0.05+-	86.53ms (85.50, 89.03) 1.16+-
master	10.73s (10.58, 10.82) 0.08+-	1.04s (1.03, 1.05) 0.01+-	763.55ms (753.46, 776.05) 7.54+-	9.93ms (9.83, 10.03) 0.07+-	85.87ms (84.85, 86.79) 0.62+-
evaluation	-0.02%, -0.0s invariant (-0.04d, 0.94p, 0.07std)	+0.38%, 0.0s invariant (0.31d, 0.58p, 0.01std)	+0.74%, 5.67ms invariant (0.32d, 0.56p, 15.53std)	-0.56%, -0.06ms invariant (-0.90d, 0.12p, 0.06std)	+0.75%, 0.65ms invariant (0.70d, 0.22p, 0.89std)
CairoMakie	12.41s (12.18, 12.64) 0.16+-	1.44s (1.43, 1.46) 0.01+-	274.49ms (271.44, 281.47) 3.69+-	12.63ms (12.42, 12.83) 0.13+-	7.06ms (6.99, 7.24) 0.09+-
master	12.43s (12.14, 12.79) 0.24+-	1.44s (1.43, 1.46) 0.01+-	272.32ms (270.43, 275.21) 1.83+-	12.58ms (12.44, 12.87) 0.14+-	7.06ms (6.94, 7.13) 0.08+-
evaluation	-0.14%, -0.02s invariant (-0.08d, 0.88p, 0.20std)	+0.00%, 0.0s invariant (0.00d, 1.00p, 0.01std)	+0.79%, 2.17ms invariant (0.75d, 0.20p, 2.76std)	+0.38%, 0.05ms invariant (0.35d, 0.52p, 0.13std)	+0.13%, 0.01ms invariant (0.10d, 0.85p, 0.09std)
WGLMakie	13.12s (12.65, 13.33) 0.28+-	1.36s (1.29, 1.43) 0.05+-	11.82s (11.51, 12.15) 0.20+-	15.71ms (13.82, 23.07) 3.28+-	1.22s (1.11, 1.46) 0.12+-
master	13.21s (12.72, 13.61) 0.29+-	1.39s (1.31, 1.49) 0.07+-	11.86s (11.47, 12.53) 0.33+-	15.01ms (13.92, 16.15) 0.90+-	1.30s (1.13, 1.46) 0.15+-
evaluation	-0.67%, -0.09s invariant (-0.31d, 0.57p, 0.28std)	-2.53%, -0.03s invariant (-0.59d, 0.30p, 0.06std)	-0.39%, -0.05s invariant (-0.17d, 0.76p, 0.26std)	+4.45%, 0.7ms invariant (0.29d, 0.60p, 2.09std)	-6.11%, -0.07s noisy🤷‍♀️ (-0.57d, 0.31p, 0.13std)

[jkrumbiegel]

What about changing display(plot; inline=false) to display(plot; window = true) instead? I never found inline that clear to be honest. Although, does opening an os window or a browser tab count as a "window"? At least, if I think about it, the default in Julia is to display however the display stack is currently configured. If I instead always want to force a window, shouldn't that be positively affirmed using a true option instead of negating the opposite?

We could also think about having a slightly more "convenient" syntax for opening new windows with this. Instead of window = true one could maybe pass window = :new or so. Just a suggestion. But I felt so far users didn't find the display(GLMakie.Screen(), fig) super intuitive.

[ffreyer]

I would also suggest adding inline as a theme argument to make it more accessible. display feels kind of niche to me in Makie and it also only applies to this one figure.

[SimonDanisch]

Yeah I think those are good ideas, but right now I just want to make a small bug fix PR.
We can target bigger API changes in the beta 0.20 branch!
The display(fig; inline=true) thing is basically also just a bug fix, considering that I intended all activate! kwargs to be usable in display too.