nextind("", 1, 0) should probably work

[StefanKarpinski]

#31312 is a fix for chop that seems to be hiding a deeper problem, namely that nextind("", 1, 0) should probably work instead of throwing an error. In general, when the number of characters to move is zero, I believe that any 0 ≤ i ≤ ncodeunits(s)+1 should be allowed in order to agree with thisind. When n > 0 then it makes sense to require 0 ≤ i < ncodeunits(s)+1. Motivation: well, the code that #31312 fixes is a perfect example. Similarly, prevind("", 0, 0) should also work.

cc @bkamins

[bkamins]

If we mentally agree that 0 and ncodeunits(s)+1 are "would be" valid indices then this is consistent.
As the design for n=0 assumed that you can call it using a valid index only and will get a valid index in return. I do not see the risk when it would hurt to extend the definition the way you propose.

When this is fixed in the docstring s should be replaced by str both in prevind and nextind (wrong variable name was used in 3 places).