Tilde ('~') character is incorrectly URL-encoded with Python 2

I'm referring to the older, unresolved issue here (it seems I cannot create a new issue nor comment on the older one over at issues.apache.org though I have a long-time user there...):

https://issues.apache.org/jira/browse/LIBCLOUD-979

I can confirm that the reported issue happens also for Amazon S3, not only for CEPH as originally reported. And the problem is indeed that libcloud URL-encodes the tilde '~' sign to '%7E', which it shouldn't according to RFC 3986.

The problem seems specific to Python 2 because in Python 3 urllib.quote() has been already updated to not URL-encode the tilde '~' sign anymore:

https://bugs.python.org/issue16285

Because of this, I think the correct fix (other than the patch proposed in LIBCLOUD-979) would be adding the tilde '~' character to the safe characters in the urlquote() utility function in utils/py3.py to ensure consistent behavior between Python 2 and Python 3.

```diff
--- libcloud/utils/py3.py 2020-04-30 12:58:55.804591828 +0000
+++ libcloud_fixed/utils/py3.py 2020-04-30 12:59:25.975368813 +0000
@@ -211,7 +211,7 @@

tostring = ET.tostring

- def urlquote(s, safe='/'):
+ def urlquote(s, safe='/~'):
if isinstance(s, _real_unicode):
# Pretend to be py3 by encoding the URI automatically.
s = s.encode('utf8')
```

Start in utils/py3.py at the urlquote() utility and compare its Python 2 behavior with Python 3 and RFC 3986. The work is done when tilde characters remain unencoded consistently across both Python versions, including the Amazon S3 case described in the issue.

Written by the indexing model from the issue text.