[BUGFIX] Fix turbine selector in random search layout optimizer

[misi9170]

While working on #919 , @paulf81 and I found a bug that meant that, during the selection of a random turbine to move in the LayoutOptimizationRandomSearch optimization routine, the num_turbinesth (last) turbine was never selected.

The issue stems from numpy's randint excluding the high value, compared to the builtin randint including the high value. I introduced the bug when I switched from importing the builtin random to importing random from numpy in 76a1465.

The fix is simply to move to the high value being num_turbines, rather than num_turbines - 1. However, to avoid confusion, I've also made the usage of numpy explicit by removing the direct import of random from numpy and instead using np.random.randint().

The following code demonstrates the difference between the builtin random.randint() and np.random.randint(). The first for loop produces both 0 and 1s, whereas the second produces only 0s.

from random import randint
import numpy as np

print("Built in")
for _ in range(20):
    print(randint(0,1))

print("\n\nnumpy")
for _ in range(20):
    print(np.random.randint(0,1))

Notes

• The test random_search_layout_opt_regression_test.py values were updated as this changed the regression test results.
• Merging this into develop will likely create a small merge conflict with Add WRG file support to FLORIS #919, where this is partially fixed. I will resolve that merge conflict once this is approved and in the develop branch.
• np.random.randint is also used in selecting initial positions in _gen_dist_based_init(). Here, the usage will mean that the (integer) maximum value will not actually be selected. However, moving 1m less than the max is essentially equivalent in this case; moreover, _gen_dist_based_init() is only used for generating initial positions that will very likely be updated by the main optimization routines. I think we don't need to change the operation of these lines, but I've still changed them to be explicit in their numpy usage after removing the from numpy import random line.