feat!: Support matrix data with units

[icweaver]

Description

Follow-up to https://discourse.julialang.org/t/juliahub-raises-65m-series-b-and-launches-dyad-3-0/136941/20

Allows for things like this to work now:

using CairoMakie, DynamicQuantities

fig, ax, p = heatmap(rand(3, 4)u"kg")

Colorbar(fig[1, 2], p; label = "Mass")
Colorbar(fig[1, 3], p; label = "Mass", dim_convert = Makie.DQConversion(us"g"))
Colorbar(fig[1, 4], p; label = "Mass", label_suffix = "unit: ({})")

fig

Changes listed here: https://github.com/icweaver/Makie.jl/blob/units-matrix/CHANGELOG.md#unreleased

Would something like this make sense to add? At any rate, thanks all for the exciting developments that are going into v0.25 and beyond!

Type of change

• Breaking change (fix or feature that would cause existing functionality to not work as expected)

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.
• Implement format and use_short_units for DQ

[ffreyer]

With this, the heatmap/image/etc. data will be forced¹ to synchronize with other plots that map to dim 3. E.g. if you add a z component to scatter, that will have to synchronize. (You can use z values for ordering/layering in Axis.) Or if you plot multiple heatmaps.
Is that something we want? Or don't want? Or don't care about?

¹ Breaking also added force_dimconverts as a plot kwarg which can be set to false to allow unitless data to pass through. So it's not entirely forced, but still what happens by default

[icweaver]

Oh cool, force_dimconverts is a really nice escape hatch:

fig = Figure()
ax = Axis3(fig[1, 1])

scatter!(ax, 1:3, 1:3, (4:6)u"cm")
heatmap!(ax, 1:3, 1:3, rand(3, 3); force_dimconverts = false)

fig

Would adding a unit-aware method for transform! be feasible too? Then, being able to do something like this seems intuitive to me:

 fig = Figure()
 ax = Axis3(fig[1, 1])

 scatter!(ax, 1:3, 1:3, (4:6)u"cm")
 heatmap!(ax, 1:3, 1:3, rand(3, 3); transformation = (:xy, 5u"cm"),  force_dimconverts = false)

 fig

I'm still exploring the internals, but it seems like more invasive changes would need to be made if we were to go with implementing a totally separate color dim convert instead of just re-using the machinery that's already in place, so I feel inclined to just stick with the above design decision and document the force_dimconverts option for folks that need it.

I rarely if ever work in 3D, so apologies if I am missing other common use cases where this might be a problem.