Closes #9091.

This PR changes the Jupyter notebook magic to store the relevant QueryJob instance on errors to the specified destination variable (if any).

How to test

Start an ipython session from the bigquery/ directory, and try to execute a query that results in an error:

In [1]: from google.cloud import bigquery
In [2]: %load_ext google.cloud.bigquery
In [3]: %%bigquery target_var --dry_run 
  
        SELECT SELECT * FROM `bigquery-public-data.samples.shakespeare` LIMIT 5; 

Expected result:
If an error occurs, the target variable is set to a QueryJob instance, allowing further introspection. This happens both in dry runs and actual runs.