missing checks for duplicate tables between translators

[Aaron Rosen]

Currently one can still create duplicate tables in a datasource driver as the check for duplicate tables is only taking into account the values of one translator at a time.

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/135099
Committed: https://git.openstack.org/cgit/stackforge/congress/commit/?id=8d7b9b8b87e034355421ad2c11e5926ed254aef4
Submitter: Jenkins
Branch: master

commit 8d7b9b8b87e034355421ad2c11e5926ed254aef4
Author: Aaron Rosen <email address hidden>
Date: Mon Nov 17 14:12:02 2014 -0800

    Refactor datasource translator logic

    Previously the method convert_obj() had a lot of logic which would
    verify that a translation table schema was valid. This code was run each time
    update_from_datasources() was called and the check only really needed to be
    done once since the translations tables do not dymanically change.

    The validation of schema is now done once when a datasource driver calls
    register_translator(translator). Having the datasource driver register the
    schema solves the problem of tables not showing up for a datasource driver
    unless it makes a full pass through update_from_datasources() which can take
    a while depending on the speed of a datasource. Now the tables show up when
    the driver loads.

    In addition, previously there was a bug where multiple translators could
    define the same table name and this would go undetected. Now we check
    for this because we keep track of these tables in the DataSourceDriver class
    direclty now.

    Closes-bug: 1393577

    Partially implements blueprint: refactor-datasource-driver-framework

    Change-Id: Iacf9144cad7c60251c138d8651cf55e331f8fe53