Add .bin_centers to BinnedStatisticResult (and possible API change)

Is your feature request related to a problem? Please describe.

1. I used binned_statistic and binned_statistic_2d quite a bit, but often need to get the bin centers out to plot (statistic) vs. (bin centers). So, in 1D, I always end up typing this:

   stat = binned_statistic(…) bin_ctr = 0.5 * (stat.bin_edges[1:] + stat.bin_edges[:-1])

2. As a separate but related point, in 1D and DD, the bin edges can be retrieved with stat.bin_edges, but in 2D, the bin edges are provided separately as .x_edges and .y_edges. I think this makes the API less consistent, and would propose that BinnedStatistic2dResult should instead have .bin_edges.

Describe the solution you’d like

1. I propose either adding a new attribute to the BinnedStatisticResult NamedTuple’s to store the bin centers, or promoting these into small classes and adding a property that computes the bin centers on the fly from the bin edges.

2. I propose deprecating .x_edges and .y_edges in BinnedStatistic2dResult and moving to .bin_edges.

Thanks for reading!

[mdhaber]

Note that we already have _make_tuple_bunch, which allows us to convert the output to an object with as many fields as we want but still works like a tuple (with the original elements) if it is unpacked.

[rkern]

The result class could implement __len__() and __getitem__() to emulate the legacy tuple items while still adding more additional attributes that are only accessible as attributes.

class BinnedStatistic2dResult(object):
    def __init__(self, statistic, x_edge, y_edge, binnumber):
        self.statistic = statistic
        self.x_edge = x_edge
        self.y_edge = y_edge
        self.binnumber = binnumber
        self.bin_edges = [x_edge, y_edge]
    def __len__(self):
        return 4
    def __getitem__(self, i):
        return (self.statistic, self.x_edge, self.y_edge, self.binnumber)[i]

One other option for bin centers might be to add a utility function that converts bin edges to the (centers, widths) representation. After the discussion about bin centers on numpy/numpy#13194, it seems that the manual conversion is more of a burden for people than I would have expected.