qcut returning confusing first bin with all positive integer series

[gerigk]

bins[0] -= 0.001 * (x.max() - x.min())
should probably be tuned - especially for integers
maybe related to issue #1462

qcut(Series([0,1,2,3,4,5,3531132324355]))

Factor:
array([(-3531132324.355, 1.5], (-3531132324.355, 1.5], (1.5, 3], (1.5, 3],
(3, 4.5], (4.5, 3531132324355], (4.5, 3531132324355]], dtype=object)
Levels (4): array([(-3531132324.355, 1.5], (1.5, 3], (3, 4.5], (4.5, 3531132324355]], dtype=object)

qcut(Series([0,1,2,3,4,5,3535))

Factor:
array([(-3.535, 1.5], (-3.535, 1.5], (1.5, 3], (1.5, 3], (3, 4.5],
(4.5, 3535], (4.5, 3535]], dtype=object)
Levels (4): array([(-3.535, 1.5], (1.5, 3], (3, 4.5], (4.5, 3535]], dtype=object)

[wesm]

This is actually very similar to the quantcut function in the R gtools package:

> quantcut(c(0,1,2,3,4,5,3531132324355), c(0, 0.25, 0.5, 0.75, 1))
[1] [0,1.5]        [0,1.5]        (1.5,3]        (1.5,3]        (3,4.5]       
[6] (4.5,3.53e+12] (4.5,3.53e+12]
Levels: [0,1.5] (1.5,3] (3,4.5] (4.5,3.53e+12]

I don't think it should be tweaking the low end, let me have a look

[wesm]

Made some fixes here and now the output matches R.