Convert matplotlib step drawstyles to plotly line shapes

[robertoffmoura]

mpl_to_plotly converts plt.step plots (and any drawstyle="steps-*" line) as ordinary straight lines, connecting the points with diagonals instead of the vertical/horizontal step segments matplotlib draws.

Before: the converted trace has line.shape = None, so plotly draws diagonal connections between consecutive points.

After: the drawstyle is mapped to plotly's step shapes, matching matplotlib's geometry exactly:

matplotlib drawstyle	plotly line.shape
steps-pre (and plt.step, steps)	vh
steps-post	hv
steps-mid	hvh

Steps to reproduce:

import matplotlib
matplotlib.use("Agg")
import matplotlib.pyplot as plt
import numpy as np
import plotly.tools as tls

x = np.linspace(0, 10, 20)
y = np.sin(x)

fig, ax = plt.subplots()
ax.step(x, y)
fig.savefig("step_mpl.png")

p = tls.mpl_to_plotly(fig)
p.write_image("step_plotly.png")

print(p.data[0].line.shape)   # "vh" with the fix; None before

matplotlib	plotly before	plotly after

[camdecoster]

Looks good! Could you please add a changelog entry? Also, I've noticed that these tests don't run in CI. Do you know if that's intentional? It's not from a recent change that you made, but I'm curious if you know the history.