Inform users of precision of different judge solutions

[Matistjati]

There have been at least ~6 problems in Kattis contests with float answers, where the judge solution has floating point inaccuracies. I think we could help reduce the risk of this by displaying the largest relative/absolute error that each submission gets relative the .ans files. Or at least warn if a particular submission is close to the error limit.

I haven't made many float problems at all, so not sure whether this a good idea. But I think it is? Summoning @Tagl since you do a lot of geometry problems, and therefore floating point problems.

[Matistjati]

Additionally, it would probably be a good idea to warn if a problem has strict precision requirements, but the answer files contain few digits.

[Tagl]

I am the number one floating point hater in my country. In my geometry problems I generally try to avoid the judge solution using floats at all unless avoiding them is extremely complicated.

I strongly agree with both suggestions.