Implement random.seed()

[OmarMesqq]

Fixes #2110.

[OmarMesqq]

@certik Can I use OS supplied PRNGs? I was thinking of writing the generator on my own, but it wouldn't be as cryptographically secure as /dev/urandom for instance

[certik]

What issue is this trying to fix?

I would use the libc rand() and also allow to initialize it.

[OmarMesqq]

@certik
it fixes the blocker reported by @rebcabin
With this final commit, running:

from random import random 
print(random())

yields a different random number every run

[certik]

Perfect, thanks! I left some comments.

[certik]

Thanks. I think this approach might work, I'll play with this soon.

[certik]

We have to be 100% compatible with CPython. In CPython, random.seed(0) initializes the seed, so the following returns the same numbers in CPython:

    random.seed(0)
    print(random.random())
    random.seed(0)
    print(random.random())

But two different number in LPython (since 0 is used for clock initialization).

[certik]

@OmarMesqq I fixed it up, let's see if this works.

[OmarMesqq]

@certik Nice! Sorry, I didn't create more thorough tests. Already paying attention to the new bug issue you opened. Thanks Ondrej!

[certik]

Thank you!

[Shaikh-Ubaid]

We are comparing floating point values here. I think we need to use range comparison abs(a - b) < 1e-5 here.