Add possibility to collect legend entries from multiple axes

[jusack]

Description

I added a convenient way to gather legend entries from multiple axes, i.e.

using GLMakie

fig = Figure()
ax1 = Axis(fig[1,1])
ax2 = Axis(fig[1,2])
lines!(ax1, randn(10), label="Trace 1")
lines!(ax1, randn(10), label="Trace 2")
lines!(ax1, randn(10), label="Trace 3")
lines!(ax2, randn(10), label="Trace 1")
lines!(ax2, randn(10), label="Trace 2")
Legend(fig[1,3], [ax1, ax2], merge=true)

will produce a joint legend for both axes.

The main reason why this is relevant for me is that the interactivity is then linked as well:

Type of change

• New feature (non-breaking change which adds functionality)

Checklist

• Added an entry in CHANGELOG.md (for new features and breaking changes)
• Added or changed relevant sections in the documentation
• Added unit tests for new algorithms, conversion methods, etc.
• Added reference image tests for new plotting functions, recipes, visual options, etc.

[jkrumbiegel]

One of the tests is really enough, given the super simple mechanism not much is gained by piping many different plot types through. Rather, it could be tested once that the legend actually has the two expected entries.

[jusack]

Thank you for the comment and looking at the PR. I have adjusted the test. However I am not sure if I the indexing into all the layers is the best way to get to the legend content. Is there an easier way to achieve what I was trying?

[jkrumbiegel]

Docstring should be updated here

[ffreyer]

This isn't really related to the pr, but shouldn't the types here be Union{AbstractAxis, Scene} so this can work with other axis types more easily? Like GeoMakie.GeoAxis or PolarAxis? (And maybe we should also have get_plots(ax::AbstractAxis) = ax.scene?)

[jkrumbiegel]

Sounds sensible

[jkrumbiegel]

Would be good to add a simple visual example here, maybe one that uses merge too

[jkrumbiegel]

two more comments got stuck in the pipeline, I've submitted them again now. The indexing in the test is fine, using internals internally is ok :)

[StefanPofahl]

I would like share my view:
My opinion about the functionality is: It is very important to add such a feature to the function Legend()!
Thanks a lot Julius for your work on this!
My proposal is:
What about an automatism: if there is a vector of axis handles join the legends of the axes?