New slice function of Histdd to support binning fragment

[zihaoxu98]

This PR modifies the function Histdd.slice, to let it support binning fragment when the start or stop are not exactly at the bin edges. The fragments are calculated by multiplying the original histogram by the bin volume inclusion fraction.

For example,

is from the the following code

from multihist import Histdd
import numpy as np
import matplotlib.pyplot as plt

sample_size = int(1e6)

x = np.random.uniform(0, 1, sample_size)
r = np.sqrt(np.random.uniform(0, 100, sample_size))
hist = Histdd(x, r, bins=[np.linspace(0, 1, 30), np.linspace(0, 10, 11)], axis_names=['x', 'r'])

hist.project(axis=1).plot(label='Full hisotgram')
hist.slice(3.2, 7.6, axis=1).project(axis='r').plot(label='Sliced histogram from 3.2 to 7.6')
hist.slice(-2, -0.5, axis=1).project(axis='r').plot(label='Sliced histogram from -2 to -0.5')
hist.slice(8.4, 12, axis=1).project(axis='r').plot(label='Sliced histogram from 8.4 to 12')

plt.legend()
plt.show()