Add Cmd.parse_args() method to parse arguments, and have Cmd.__call__()
take parsed argument Namespace object as argument.

get_func returns initialized Cmd object

The library now handles various authentication failures automatically, so
the Reset command is no longer necessary.

Better error messages for Kerberos failures.

This cleans up version strings, since unknown versions should just resolve
to the unversioned document

This is the unfortunate consequence of the authentication mechanism not
providing standard HTTP authentication failures that coucld be handled
entirely in HttpFetcher.  The results from the query have to be parsed to
determine if the authentication has failed.

This allows users of the fetcher to call this method explicitly if needed.

It's unfortunate that the authentication mechanism forces the callers to
parse returned pages to know if authentication has failed.  This should be
handled entirely inside of the fetcher without the caller having to worry
about it at all.

This new method also has some smarter error handling.
In case of ecp-cookie-init failure, klist will be called to determine if
lack of kerberos principle is the failure, and an appropriate exception will
be thrown.

add _build_dcc_url() method to make it simpler to build arbitrary dcc urls.

and update for python3

This commit contains fixes to the code to allow it to be interpreted by
Python 3.5. It also contains some changes to the code to move the cookie
authentication routines to comms.py and away from the command line
interpreter, prompted by some of the cookie code being incompatible
with Python 3.5.

This commit contains code to parse the abstract, notes, keywords, publication info and journal reference fields contained
within a DCC record. These are accessible as fields within a DccRecord object.

This commit removes the majority of the HTML parsing used previously to extract information
from DCC documents. This was fragile and reliant on regular expressions and was likely to
cause more future problems as the DCC is altered. It has been replaced by the XML parser
committed previously to this branch.

Some HTML parsing is still used to identify when an error occurs. This is needed because
the DCC will produce an HTML page in the event of an error, even if an XML document has
been requested.

In addition, internal document IDs are now recognised for "referenced by" and "reference
of" fields, though ideally these should be DCC numbers (we have requested this from the
developers).