Color levels not respected

[WesleyTheGeolien]

ALL software version info

holoviews 1.13.5 pypi_0 pypi
Python 3.8.5 (default, Sep 4 2020, 02:22:02) [Clang 10.0.0 ] :: Anaconda, Inc. on darwin

Description of expected behavior and the observed behavior

When passing colour_levels to an image opts color_levels I would expect them to be respected eg. the colormap is broken on the limits that are specified.

As you can see in the example below the color_levels specifies 0 as a limit however the purple step includes values greater than 0. (In the data provided this is off the eastern African coast)

Complete, minimal, self-contained example code that reproduces the issue

data here: https://github.com/CEREGE-CL/netcdf_editor_app/blob/main/data/standard.nc

import xarray as xr
import holoviews as hv
import numpy
hv.extension('bokeh')

ds = xr.open_dataset('data/standard.nc')
var = 'Z'

abs_max = numpy.abs((ds[var].min(), ds[var].max())).max()

delta = 1000
clevels = [-abs_max] + [*numpy.arange(0, abs_max, delta)[::-1] * -1, * numpy.arange(0, abs_max, delta)[1:]] + [abs_max]
print('color_levels: ', clevels)

img = hv.Image(
    ds[var],
    group="Map",
).opts(
    invert_axes=True, tools=['hover'], width=600, aspect=2, colorbar=True, cmap='Category10', clim=(-abs_max, abs_max), color_levels = clevels
)

img

Screenshots or screencasts of the bug in action

[WesleyTheGeolien]

So it looks like this function is the culprit:

holoviews/holoviews/plotting/util.py

Line 924 in 8e137c0

def color_intervals(colors, levels, clip=None, N=255):