Add triangular distribution and related tests.

[mvernacc]

Add the triangular distribution to the set of available uncertainties.

Example usage of the new feature:

"""Demonstrate new trianguar distribution feature in Rhodium."""

import numpy as np
import matplotlib.pyplot as plt
from scipy import stats
import rhodium as rdm
import seaborn as sns

# Triangular distrib properties
min_value = 0; max_value = 1; mode_value=0.25
scale = max_value - min_value
c = (mode_value - min_value) / scale


def identity(x):
    return x

model = rdm.Model(identity)

model.parameters = [
    rdm.Parameter('x')
]

model.responses = [
    rdm.Response('y'),
]

model.uncertainties = [
    rdm.TriangularUncertainty('x', min_value, max_value, mode_value)
]

scenarios = rdm.sample_lhs(model, nsamples=1000)

sns.distplot(scenarios['x'])

# Plot the pdf
x = np.linspace(min_value, max_value)
density = stats.triang.pdf(x, c=c, scale=scale, loc=min_value)
plt.plot(x, density, linestyle='--', color='black', label='True PDF')

plt.title('Samples from a Triangular Distribution')
plt.xlabel('Value')
plt.ylabel('Density')
plt.legend()
plt.show()

[dhadka]

Merged in #37