Off by one bug in digit_bad test

[patricksurry]

Noticed while messing with word headers. The digit_bad constant is defined as six characters via s" /:@[{" ( addr u ) drop constant digit_badbut the test was iterating from 0..6 inclusive, so the last value was actually the first byte of thedigit_badheader, i.e. the word length of 9 which was showing as a tab inresults.txt`. Reduced loop size by one.

The unrelated never-native difference that's shown is from a previous PR, must not have updated the test output for that.

[SamCoVT]

That's a good catch - I doubt I ever would have found it.