Fix the behavior of negative zero

[LighghtEeloo]

Hi, I tried to compile and pass all the tests, but one failing test caught my eye:

---- test::string_of_negative_zero_is_zero stdout ----
thread 'test::string_of_negative_zero_is_zero' panicked at 'assertion failed: `(left == right)`
  left: `"0"`,
 right: `"-0"`', src/lib.rs:545:9

It reminds me of the signed zero in IEEE 754, which indicates the sign of a zero, regardless of the fact that 0.0 == -0.0. Therefore, a direct to_string() will make the test fail if the test case constructs -0.0 deliberately.

This patch passes the test by additionally judging whether the number to be returned equals zero. Since 0.0 == -0.0, both should be treated as 0.0 when printing.

It's is a tiny patch and only takes perhaps less than a minute to check and comment. The library is great, I find it handy. Thanks, and have a nice day / night : )

[glebm]

@shepmaster Can you please merge this PR to fix the failing test?

[shepmaster]

Interesting; this is standard library behavior that changed somewhere along the way:

echo 'fn main() { println!("{}", -0.0); }'  > a.rs
rustc +1.48 a.rs 
./a 
# => 0

rustc +1.56 a.rs
./a
# => -0

[shepmaster]

Thanks! The tests still fail, but for internal reasons, unrelated to this PR.