Sample to the bounds in quasi random optimizers

[jrapin]

Types of changes

• Docs change / refactoring / dependency upgrade
• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to change)

Motivation and Context / Related issue

Ugly patch for #515
Better methods should be implemented at some point :s

How Has This Been Tested (if it applies)

Checklist

• The documentation is up-to-date with the changes I made.
• I have read the CONTRIBUTING document and completed the CLA (see CONTRIBUTING).
• All tests passed, and additional code has been covered with new tests.

[jrapin]

Illustrations are based on the following snippet:

import nevergrad as ng
import numpy as np
budget = 100
parametrization = ng.p.Tuple(*(ng.p.Scalar(lower=-12, upper=12) for _ in range(2)))
optimizer = ng.optimizers.LHSSearch(parametrization, budget=budget)
points = np.array([optimizer.ask().value for _ in range(budget)])
%matplotlib
import matplotlib.pyplot as plt
plt.plot(points[:, 0], points[:, 1], 'x')
plt.gca().set_xlim([-12, 12])
plt.gca().set_ylim([-12, 12])

With the current master, this provides:

After this PR, this would sample as follows:

With the following tag we can get mixed behavior (sampling on the full range on x, and according to the distribution on y:

parametrization[1].full_range_sampling = False

[jrapin]

discussed with @teytaud and accepted