py3: fix one doctest in French book

[chapoton]

by marking it as random

[chapoton]

New commits:

​f81acd9

py3: mark one doctest as random in the French book

[chapoton]

please review

[chapoton]

green bot, please review

[tscrim]

I don't really like marking tests as random when there is exact output. I think the better thing to do is to change the test to check that the output matches something explicit.

[chapoton]

Well, I have not read the book, but this precise doctest is probably here to display an example of bad answer, by lack of sufficient precision...

[chapoton]

page 201 of the book (​http://sagebook.gforge.inria.fr/english.html)

Let us continue. For the following system, J.variety() computes (exactly) a
triangular decomposition of J, then finds numerically the real solutions of the
obtained system(s). This yields a unique real solution:

sage: R.<x,y> = QQ[]; J = ideal([ x^7-(100*x-1)^2, y-x^7+1 ])
sage: J.variety(RealField(51))
[{y: 396340.890166545, x: -14.1660266425312}]

Yet, by performing the computation exactly until the end, we see there are three
real solutions, and the value of x in the above numerical solution is completely
wrong:

[tscrim]

Based on that, I am more convinced that we should make sure it is testing the correct output and not just something from a valid computation. However, there is something very annoying happening when I try to test against the value exactly. The x value does not quite match despite having the same digits displayed. So I guess we just have to let this be with # random. :/