Fix double '//' on URLs with empty authority and '//'-prefixed path (#185)

[deepakganesh78]

Fixes #185.

Root cause

furl('////path').url returned '//////path' instead of '////path' on Python 3.9+.

In FurlBase.tostr(), when scheme is None and netloc == '', furl unconditionally prepended '//' to denote the empty authority:

if self.netloc == '':
    if self.scheme is None:
        url = '//' + url

Since Python 3.9, urllib.parse.urlunsplit() already prepends '//' to a result whose path begins with '//' (per RFC 3986 §3.3, to prevent the path from being misparsed as an authority). furl then added a second '//', doubling the leading slashes. This is the stdlib change referenced in the issue and is why the existing test_odd_urls assertion (furl('////path').url == '////path') began failing on 3.9+.

Fix

Only prepend '//' when urlunsplit()'s output doesn't already start with it. All other cases are unchanged; '////path', '///', '///x', '//', '////' now all round-trip correctly.

Tests

Extended test_odd_urls with round-trip assertions for '////path', '//', '///', '///x', '////'. Added a changelog entry.

Full suite: 75 passed, 2 failed. The 2 failures (test_hosts, test_netloc) are pre-existing and unrelated — they concern IPv6 validation under newer ipaddress (issue #182 / #176) and fail identically on a clean checkout without this change.