download_strategy: fix bad unescaping in basename parsing

[Bo98]

Fixes the core issue causing #15235.

• Don't erroneously run URI.decode_www_form_component over the whole URL - only run that on query string parts and run other URI unescape methods on the path part
• Don't rely on error-prone regexes or Pathname ascending when we can split it properly into an array from the URI parser

[apainintheneck]

This seems to be a good improvement. Make sure that all the uses of AbstractFileDownloadStrategy#parse_basename will work with the new method though.

I wrote a small script to compare the old and new #parse_basename results and they're pretty similar. All of the differences seem to make sense too. It seems like it fixes the following problem too.

Before	After
file=lft-3.91.tar.gz	lft-3.91.tar.gz
file=libnatpmp-20150609.tar.gz	libnatpmp-20150609.tar.gz

• Script
• Results

I wonder if it makes sense to add some tests for this method as well.

[MikeMcQuaid]

Great work so far, thanks @Bo98!

[Bo98]

I wrote a small script to compare the old and new #parse_basename results and they're pretty similar. All of the differences seem to make sense too. It seems like it fixes the following problem too.

Thanks!

Looks like no regressions here which is good to hear.

[MikeMcQuaid]

Thanks again @Bo98!