Race condition in metadata extraction

[candleindark]

While race conditions relating to accessing local dataset cache and the RepoUrl table have been addressed properly in #238, an extremely unlikely race condition is still possible in the extraction of mete-data through the execution of the extract_ds_meta Celery task. Should this race condition occurs, two or more URLMetadata records with the same extractor_name and dataset_version can be written to the database, creating duplication of data in the database.

Provide a solution to prevent this race condition or eliminate the affect of affect of it. Try to avoid using locks in the solution for more locks make dead locks more likely.