Parsing fails on Windows when filename is not relative path in current working dir

[rly]

1. On Windows, the following:

   import rnc2rng
   rnc2rng.load('citeproc/data/schema/csl.rnc')

   results in error: [Errno 2] No such file or directory: 'csl-terms.rnc'

This appears to be due to changes in #18, specifically: https://github.com/djc/rnc2rng/blob/d5405e85f8889ed54d13bfc494e81b7ffd337103/rnc2rng/parser.py#L215-L220

s.path is set to the directory name of the input filename. Windows paths include a ":", so line 216 treats the path as a URL and calls urljoin on the path. This fails to join a Windows directory name with a filename. Execution should instead enter the else statement.

2. On Windows, the following:

   import rnc2rng
   rnc2rng.load('C:/Users/Ryan/Documents/citeproc-py/citeproc/data/schema/csl.rnc')

   fails with urllib.error.URLError: <urlopen error unknown url type: c>.

Again, this is because absolute paths on Windows contain a ":" and this line: https://github.com/djc/rnc2rng/blob/d5405e85f8889ed54d13bfc494e81b7ffd337103/rnc2rng/parser.py#L693

treats the path as a url. Execution should instead enter the else statement.

[djc]

Sounds like you got a handle on this! Could you do a PR?