Make some improvements in canonicalize

[boolean5]

The discussion that took place while writing the unit tests for canonicalize highlighted several improvements that can be made. Specifically:

• Use urlparse to extract the hostname and the path from the url instead of using a regular expression. This is expected to simplify the code significantly.

• Add a docstring that will mention the differences between our implementation and the one suggested in the Safe Browsing v2 docs.

• Add a couple of extra test cases:

  • A test case that checks that escape sequences for the tab, CR and LF characters (i.e., %09, %0d and %0a) are not removed (this is one of the canonicalization rules mentioned in the Safe Browsing v2 docs but it's not covered by the suggested test cases).

      ("retain_escaped_tab_cr_lf", "http://www.google.com/foo%09bar%0dbaz%0a2", "www.google.com/foo%09bar%0Dbaz%0A2"),
    

  • A test case that checks that a url is canonicalized correctly when it contains a
    username and password.

• Remove a few TODO comments that don't seem to be important any more (the ones that suggest using d, _subs_made = re.subn(...) instead of d = re.subn(...)[0]: 1, 2, 3).

• I would also suggest renaming the d parameter of canonicalize to domain or url. Avoiding single-letter variable names makes for more readable code.