Incorrect lower bound from StepBy<A, RangeInclusive<A>>::size_hint

[scottmcm]

Repro: https://is.gd/nFIian (tried 2017-04-21)

#![feature(step_by)]
#![feature(inclusive_range_syntax)]

fn main() {
    let i = (0...10).step_by(3);
    assert!(i.size_hint().0 <= i.count());
}

Looks like the implementation is adding one to the Range (exclusive) lower bound, which is incorrect when the size of the base range isn't a multiple of the step.

Thoughts on whether it'd be a better fix to just not add one (correct, but a less-helpful hint) or to add one to end before calling Step::steps_between (better hint, more math and more overflow cases)?

[cuviper]

From what I saw on #42514, the upper bound can also be one too high. That's less problematic, I suppose, although the proposal to "just not add one" should not apply to the upper bound since then it would be too low in some cases.

Of course, it's a bit moot since this is now deprecated and expected to be removed soon. :)