Fix documentation of fns rand_core::le::read_u*_into

[dhardy]

Fixes some documentation, as pointed out in #1624.

I considered deprecating or renaming these fns, but they have some utility and only really need doc fixes.

[newpavlov]

What about changing the asserts to equality?

[dhardy]

I would, but that's technically a breaking change.

[newpavlov]

Technically, yes. But I think it's fine since I don't think any downstream users rely on the non-equal case and we could declare that it's a "bug fix" (if we are to stretch the meaning far enough). We had a similar "breaking change" in getrandom: rust-random/getrandom#614.

[vks]

Why do you want to change the assert? I don't really see the harm, any unused source bytes are just ignored.

[dhardy]

Unused input bytes are a possible indicator of mis-use. But I think it's best not to change the assert now (possibly on the next breaking change, but not important).