sample_size not being casted to integer

Hi, I have come across this problem while using the library.

Here, sample_size is not being casted to integer after round, resulting in sample_size being a float:

https://github.com/coin-or/rbfopt/blob/8379229de40721cd1160869f602eb126c8b3d552/src/rbfopt/rbfopt_utils.py#L464-L465

This causes trouble down the line as np.random.permutationrequires an integer (it raises IndexError: tuple index out of range when it's a float):

https://github.com/coin-or/rbfopt/blob/8379229de40721cd1160869f602eb126c8b3d552/src/rbfopt/rbfopt_utils.py#L305-L306

Here is a minimal code example that triggers an error due to sample_size being float.

import rbfopt
import numpy as np

def obj_funct(x):
    return x[0]*x[1] - np.sum(x[2:])

types = np.array(['R', 'R', 'I', 'I'])
dim = len(types)
lB = np.array([0] * dim)
uB = np.array([10] * dim)

bb = rbfopt.RbfoptUserBlackBox(dim, lB, uB, types, obj_funct)
settings = rbfopt.RbfoptSettings(max_evaluations=50)
alg = rbfopt.RbfoptAlgorithm(settings, bb)
val, x, itercount, evalcount, fast_evalcount = alg.optimize()

I hope that this helps to replicate the error (I'm using Python 2.7 and rbfopt installed through pip).

Regards, Javier.

coin-or / rbfopt

sample_size not being casted to integer #21