randint() implementation is flawed

[goualard-f]

Describe the bug

The randint() function in the Calculation module uses a flawed algorithm that can make it return values outside the requested domain. This can happen in two ways:
1/ When providing bounds that lead to an integer overflow;
2/ When the underlying floating-point RNG returns 1.0f (in the Random::random<float> implementation, there are 128 integers out of 2^{32} that can give that result ---approx. 1 in 3e8).

Relying on Random::random<float> also entails that randint() cannot ensure uniformity: Ion::random() generates
integers in [0,2^{32}-1] with a uniform distribution. Once divided by 0xffffffff, some floating-point numbers are created more frequently than others due to rounding (See https://hal.science/hal-02427338). This translates to some integers returned more often than others with randint().

Proposed cure: change the algorithm used entirely (see, e.g., Lemire's nearly divisionless algorithm: https://lemire.me/blog/2019/06/06/nearly-divisionless-random-integer-generation-on-various-systems/).

Screenshots

To Reproduce

Steps to reproduce the behavior:

1. Go to the 'Calculation' app
2. Type 'randint(0,2^31-1)'
3. See error

This is the way to reproduce the first problem. The second problem, linked to the floating-point RNG cannot easily be reproduced with either the device or the simulator, since it requires calling randint() an unknown number of times.

Expected behavior

A clear and concise description of what you expected to happen.

randint(a,b) should return an integer in [a,b], not outside of it.

Environment

• Epsilon version (Settings > About > Software version).
  20.4.0
• The platform(s) on which the problem happens: online simulator, actual device, etc...
  Online simulator and actual device.