Float to_i should raise a FloatDomainError when NaN

[stevegeek]

Float to_i should raise a FloatDomainError when NaN

There is no test for this in the to_i example in spec/core . I added one and will then try upstream it to ruby/spec

I notice that the change recommended for the stack allocated Float has been fixed by a similar commit so did not include it

It was:

auto significand = Value::floatingpoint(std::frexp(arg, &exponent));
Value x = significand.send(env, "*"_s, { y })->as_float()->to_i(env);

vs what is now in master

[stevegeek]

Also opened ruby/spec#958