Probability plot missing data when using 'prob' option

Nobody has claimed this yet.

* Python version: 3.7.6
* numpy version: 1.19.2
* matplotlib version: 3.3.2
* mpl-probscale version: 0.2.5
* Operating System: Windows 10
Running in jupyter notebook.

### Description
I generated 100 data points (norm.rvs) and plotted the results using probscale. I've got two plots -- one uses 'prob' and the other uses 'qq' for the y-axis. The plot that uses qq shows all the data, while the plot that uses prob does not show all the data.

### What I Did
```python
fig, (ax1, ax2) = pyplot.subplots(figsize=(12,6), ncols=2, sharex=False)
common_opts = dict(
probax='y',
datascale='linear',
datalabel='',
scatter_kws=dict(marker='o', linestyle='none')
)

df2 = norm.rvs(0,1,size=100)
fig = probscale.probplot(df2, ax=ax1, plottype='prob', bestfit=False, problabel='Probability', color='xkcd:ocean green', **common_opts)
fig = probscale.probplot(df2, ax=ax2, plottype='qq', bestfit=False, problabel='Standard Normal Quantiles', color='xkcd:blue gray', **common_opts)
```

### What else?
If I set the y limits for the 'prob' case to something wider than the defaults, I can see the data points that were being left out.

Thanks,
Joe

Start by running the provided Python example in a Jupyter notebook with mpl-probscale 0.2.5, comparing the prob and qq plots and their default y-limits. Trace the plotting path used by probplot for plottype='prob'; done means all 100 generated points are visible without manually widening the y-axis limits.

Written by the indexing model from the issue text.