Disable Datashader count aggregator self-intersect by default for curves

[droumis]

Describe the solution you'd like

The self-intersect option for a Datashader count aggregator controls whether pixel-crossings of the same line contribute toward a pixel's count.

It was decided that a more reasonable default behavior would be to disable self-intersect by default.

This only makes sense for curves, not paths; self-intersections of paths are usually valuable information.

Additional context

import pandas as pd
import numpy as np
import holoviews as hv
from holoviews.operation.datashader import rasterize
import datashader as ds

hv.extension('bokeh')

self_intersect = True
n_samples = 1000
data = np.random.randn(n_samples)
df = pd.DataFrame({'time': np.arange(len(data)), 'value': data})

hv_curve = rasterize(hv.Curve(df, 'time', 'value'), aggregator=ds.count(self_intersect=self_intersect)).opts(width=600)

hvplot_curve = df.hvplot.line('time', 'value', rasterize=True, aggregator=ds.count(self_intersect=self_intersect))

hv_curve + hvplot_curve

Screen.Recording.2023-11-02.at.4.56.59.PM.mov

[droumis]

Also, in the video in the description, it appears that there remains some gradient in the line even when self-intersect = False. Is that right? If self-intersections are not counted and there is only one line, then all the counts should be the same and therefore I would expect only a single hue