Currently, the code to extract a query name to show in the
queries panel, expects the incoming query to always have
the `document.loc.source` attribute available. If this
attribute isn't available, an exception is thrown, and the
queries panel is displayed as a blank page. These changes
adjust the logic used to extract a query label to be more
flexible. The query name is now derivd by attempting
the following (in order):

1. Attempt to parse and extract a query name from
an incoming query string.
2. Check to see if the incoming query has already
been parsed; if so extract the query name from the
parsed document (e.g. happens when using `graphql-tag`).
3. If the incoming query has already been parsed, but
the query name can't be extracted from the result, see
if the document includes the original query source. If
so, try re-parsing it, and re-extracting the query name.
4. If all else fails, fallback on using the query
ID as the query name.

Fixes #137.