Add consistent legend support across backends

[Advaitgaur004]

Description

This PR implements consistent legend support across all plotting backends (matplotlib, bokeh, and plotly) and adds automatic legend generation for plot_psense_dist.

Key Changes

• Added full legend support for Plotly backend
• Enhanced Bokeh legend with automatic placement logic
• Added automatic legend generation for:
  • plot_psense_dist (for alpha parameter)

Implementation Details

• Implemented legend() function for Plotly backend
• Enhanced Bokeh's legend placement with auto-positioning
• Added automatic legend generation in plot_psense_dist for power scaling factor
• Ensured consistent behavior of pc.add_legend("dim_name") across all backends

Checklist

• Added/updated tests
• Updated documentation
• Added example notebooks (if applicable)
• Passes all CI checks

Issue Targeted

#109

---

📚 Documentation preview 📚: https://arviz-plots--115.org.readthedocs.build/en/115/

[amaloney]

Add more context in the linking issue, and why these features are needed. Also add visuals that explicitly show how this new code is adding to the user experience.

[Advaitgaur004]

I have added a legend placement threshold in the Plotly backend legend, which can now be altered, eliminating the need for hard coding.
And for MRE that links to bokeh.backend.legend

if side == "auto":
        plot_width = target_plot.width
        if plot_width >= legend_placement_threshold:
            side = "right"
        else:
            side = "center"

can be reproduced by following the code.

import numpy as np
from bokeh.plotting import figure, show
from bokeh.layouts import gridplot

def create_example_plot(width):
    p = figure(width=width, height=300)
    x = np.linspace(0, 10, 100)
    grid = gridplot([[p]], merge_tools=False)
    p.line(x, np.sin(x), line_color="red")
    p.line(x, np.cos(x), line_color="blue")
    legend(grid, 
           [{'color': 'red'}, {'color': 'blue'}], 
           ['sin(x)', 'cos(x)'], 
           side='auto')
    
    return grid
narrow_plot = create_example_plot(300)
wide_plot = create_example_plot(800)
show(gridplot([[narrow_plot, wide_plot]]))

300x300 dim

800x300 dim(Enough space is there so legend can be placed outside)

[Advaitgaur004]

Any Thoughts on aliasing in various backend plotting modules, one has to make thing easier for developer too @aloctavodia

[Advaitgaur004]

gentle reminder @amaloney

[amaloney]

As a user I would expect the following.

• If I request an image size of 300x300 pixels then I should get an image of 300x300 pixels.

I get this with the first example

but I do not get this with the second example.

The second plot has a total dimension of 300x800 pixels where a portion of that image is where the plot exists plus the size of the legend. This is not what I would have expected the output to be, since I requested an plot of 300x800 pixels in size.

There unfortunately is no easy solution to this.

We could instead have an argument where the user can request the legend be outside the plot area. Then in the docstring of the method we indicate that if this argument is selected, the resulting plot area will be smaller in size to accomodate the legend, given the requested figure size. How about you try implementing this idea.

[Advaitgaur004]

First of all, there is a slight error; the dimensions should be 800x300, as the height is predefined to be 300 in my test case. Therefore, it is not 300x800 but rather 800x300.

If the legend_in_place argument is set to False (indicating that the user wants the legend to be outside the plot), I will reduce the original plot width to 80% and adjust the legend to occupy the remaining 20%. Should I go this way or should we automate this thing, if the width is too large then place the legend inside the plot(reasonable) and if small then legend should be outside

I have changed the magic number to 500 for the time being and these are the test case

def create_example_plot(width):
    p = figure(width=width, height=300)
    x = np.linspace(0, 10, 100)
    grid = gridplot([[p]], merge_tools=False)
    p.line(x, np.sin(x), line_color="red")
    p.line(x, np.cos(x), line_color="blue")
    legend(grid, 
           [{'color': 'red'}, {'color': 'blue'}], 
           ['sin(x)', 'cos(x)'], 
           side='auto')
    
    return grid
narrow_plot = create_example_plot(450)
middle_plot = create_example_plot(550)
wide_plot = create_example_plot(800)
show(gridplot([[narrow_plot, middle_plot, wide_plot]]))

for 450x300

for 550x300

for 800x300

For Comparison

If seems correct I'll commit right away

[Advaitgaur004]

@amaloney Review