much better `squeeze-unhex-and-decode-utf8-string'

much better `squeeze-unhex-and-decode-utf8-string'

It turns out that `url-unhex-string' uses the multiple-concat strategy
that I just removed from `squeeze-update-state', which ends up scaling
quadratically in the size of the input.  This is bad, particularly
when we are working (as I am) with large output string such as
resulting from an albums query.