Inconsistent results for cast::i64(-0.1f64) and cast::u64(-0.1f64)

[ollie27]

Using 0.2.1 cast::i64(-0.1f64) returns Ok(0) but cast::u64(-0.1f64) returns Err(Underflow). I would expect the latter to also return Ok(0) as float to int conversions truncate towards 0 and 0 can be represented in a u64 just fine.