Suspected implementation bug in index_from_time(sample_rate, span::TimeSpan)

[ericphanson]

The implementation of the semantics isn't correct when the right endpoint of the span doesn't fall exactly on a sample; we always throw away the right-endpoint even when it is well within the span:

TimeSpans.jl/src/TimeSpans.jl

Line 237 in eb8b794

j = i == j ? j : (j - 1)

(unless that would make an empty interval, in which case we keep it).

@kleinschmidt's suggestion

I think the solution might be to convert the final index back to time, and see if it needs to be adjusted down by 1 (e.g., if time_from_index(rate, index_from_time(rate, stop(span))) == stop(span) then you use j-1, otherwise use j

would fix that, I think.

Originally posted by @ericphanson in #26 (comment)