Adds denselines recipe

[ConnectedSystems]

Description

Adds a new DenseLines recipe for visualizing time series data as a density line chart, based on the Moritz and Fisher method. It calculates and displays the density of overlapping lines, providing insights into data distribution over time.

Type of change

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• 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.

[ConnectedSystems]

Denselines implements the method of Moritz and Fisher (2018).
It plots, in essence, heatmaps of time series.

Closes #5427

@jkrumbiegel suggested that this be part of the Datashader recipe, however I opted to keep this separate at least for now.

Usage example:

using Dates
using Statistics

using WGLMakie

function example_data()
    n_times = 100
    n_series = 10000
    t = range(0, 2π, n_times)

    time_series = zeros(n_times, n_series)

    # First group: constant frequency sine wave
    for i in 1:div(n_series, 2)
        time_series[:, i] = 2.0 .+ sin.(t) .+ 0.1 .* randn(n_times)
    end

    # Second group: increasing frequency and amplitude
    for i in (div(n_series, 2)+1):n_series
        freq = range(1, 10, n_times)
        amp = range(0.5, 2, n_times)
        time_series[:, i] = 1.0 .+ amp .* sin.(freq .* t) .+ 0.1 .* randn(n_times)
    end

    return time_series
end

f, ax, sp = denselines(example_data(); bins_x=200, bins_y=100, colormap=:plasma)
ax.xlabel = "Time"
ax.ylabel = "Value"

[ffreyer]

Couldn't we also skip the pixels vector here and directly write to the matrix? That should save quite a bit of GC time I'd imagine.

Similarly, compute_normalized_density could probably clear and write to a matrix created once in denseline_data instead of creating a new one for every set of points

[jkrumbiegel]

I still think it might make sense to reuse the datashader infra for this if at all possible

[ConnectedSystems]

I've made changes for performance.

@jkrumbiegel , is a barrier function all you have in mind when you say "reuse the datashader infra"?

EDIT to add before and after:

Before: 1.797718 seconds (2.06 M allocations: 1.673 GiB, 29.31% gc time)

After: 0.377040 seconds (60.37 k allocations: 3.668 MiB)