The artistfilter module no longer fails if this json file doesn't exist

Thanks for looking into this. I opened a ticket to explain the reasoning behind Artist filtering: AB-301

The filtering process was very slow when I initially implemented this feature, so I added this quick hack with the cache file to make it faster (when I was using the data for a project).
We haven't used artist filtering since, so I think we could clean this up. I recommend we do two things:

1. RIght now, change the loading of this file to set `rtoajson = {}` if the file `recordingtoartistmap.json` doesn't exist. This is an easy fix which we should have had in place a long time ago
2. Use our built-in cache (right now Memcache, but Redis once we merge this PR: https://github.com/metabrainz/acousticbrainz-server/pull/225). This should let us look up some data mapping recording->artist and keep it in the cache so that this step gets faster.

@Alastair Porter 

While debugging the issue I found that the file artisitfilter.py is using the line

rtoajson = json.load(open("recordingtoartistmap.json"))

to build a cache for the searched mbid's . The file recordingtoartistmap.json  never gets written though apart from not getting created.  The code to fix this can be added but there is an additional issue. When I try to evaluate a dataset using "Filter by Artist" I get an error as below :

sqlalchemy.exc.ProgrammingError: (psycopg2.ProgrammingError) column "mbid" does not exist

LINE 1: SELECT mbid::text, data->'metadata'>'tags'>'musicbrainz_ar...

 So now I am not clear that Filter by artist feature is used or not. The job remains in status "running" forever when the error occurs.