Add real legend_cols support for Bokeh

[hoxbro]

In Bokeh 3.1 support for ncols and nrows was added. This PR implements that. Before that legend_cols was only used to determine if the legend should be horizontal or vertical.

Current behavior

Behavior after this PR

Matplotlib behavior

import holoviews as hv
import panel as pn

hv.extension("bokeh")
# hv.extension("matplotlib")


curves = [hv.Curve([i] * 10, label=str(i)) for i in range(10)]
data = [(0, 0, "A"), (1, 1, "B"), (2, 2, "C")]
optss = [
    dict(show_legend=True, legend_cols=2),
    dict(show_legend=True, legend_cols=2, legend_position="top"),
    dict(show_legend=True),
    dict(show_legend=True, legend_position="top"),
]


plots = []
for opts in optss:
    ol = hv.Overlay(curves).opts(**opts)
    bar = hv.Points(data, vdims=["y", "color"]).opts(color="color", **opts)
    plots.append(pn.Column(ol, bar))

pn.Row(*plots).servable()

[codecov-commenter]

Codecov Report

Merging #5669 (0408a24) into main (0b837fc) will decrease coverage by 0.01%.
The diff coverage is 63.63%.

@@            Coverage Diff             @@
##             main    #5669      +/-   ##
==========================================
- Coverage   88.29%   88.29%   -0.01%     
==========================================
  Files         302      302              
  Lines       62541    62545       +4     
==========================================
+ Hits        55220    55221       +1     
- Misses       7321     7324       +3     

Impacted Files	Coverage Δ
holoviews/plotting/bokeh/element.py	88.11% <63.63%> (-0.10%)	⬇️

... and 1 file with indirect coverage changes

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[philippjfr]

Very, very excited to finally have support for this one. The changes look good.