Boolean masks should become their own type

[Lokathor]

some people like the increased type safety.

this is a breaking change so it'd be in 0.6, but otherwise it's not hard.

[Lokathor]

Note: This will break because all existing cmp_eq operations would have to change from returning Self to returning a mask value.

[Lokathor]

Probably we can just have the "mask" factor be a wrapper type over a SIMD type, with blend / bit functions.

pub struct Mask<T>{ inner: T }

[Lokathor]

Looks like on i586 and other locations without specific hardware support for bit ops on floats we'll have to store these masks as unsigned arrays so that the bits are preserved properly. Otherwise signaling nan values can get bit modified to become quiet nan values.

I wanted to put off any breaks since the request was supposed to be "nice but not needed", but actually this only increases the priority of this particular break.