HeadTailBreaks RecursionError still not fully resolved

[martinfleis]

I have seen the RecursionError coming from HeadTailBreaks again. This time caused by a floating point imprecision.

This snippet allows you to reproduce the issue. The parquet is just 2kb containing only the problematic bit of the values.

df = pandas.read_parquet("https://www.dropbox.com/s/p9pgg2pdvnhvgsw/sample.parquet?dl=1")
bins = mapclassify.HeadTailBreaks(df["values"])

The current workaround is to round the values before sending them to mapclassify but we should somehow resolve this under the hood (not that I know how from the top of my head).

bins = mapclassify.HeadTailBreaks(df["values"].round(6))

xref #46 and #95