Plot element variables with Plots.jl

As discussed on Slack, it would be nice to be able to use Plots.jl or Makie to natively visualize element variables.

From @ranocha, the following can be used to visualize element variables over a uniform `TreeMesh`. However, it would be nice to enable more general visualization.

```julia
function get_element_variables(u_ode, semi)
element_variables = Dict{Symbol, Any}()

Trixi.get_element_variables!(element_variables, u_ode, semi)

if semi.mesh isa TreeMesh
cell_ids = semi.cache.elements.cell_ids
x = semi.mesh.tree.coordinates[1, cell_ids]
y = semi.mesh.tree.coordinates[2, cell_ids]
element_variables[:x] = x
element_variables[:y] = y
end

return element_variables
end

function do_cool_stuff()
# ...

element_variables = get_element_variables(integrator.u, semi)
hm_ind = heatmap!(ax_ind,
element_variables[:x],
element_variables[:y],
element_variables[:indicator_shock_capturing],
colorrange = (0.0, alpha_max),
colormap = :inferno)
Colorbar(fig[1, 2], hm_ind; vertical = false,
label = "Shock capturing indicator",
width = Relative(0.85))

# ...
end
```

Start with Trixi.get_element_variables! and the shown TreeMesh coordinate handling, then compare how Plots.jl or Makie could consume element variables on nonuniform meshes. Done means element variables can be visualized natively beyond the uniform TreeMesh case, with the supported generality and verification approach agreed.

Written by the indexing model from the issue text.