Logging table should be periodically truncated

[audiodude]

The logging table is set up to grow indefinitely. Recently we had an outage because it got too full:

+--------------------+---------------------------------------+------------+
| Database           | Table                                 | Size in MB |
+--------------------+---------------------------------------+------------+
| enwp10_prod        | logging                               |   42164.95 |
| enwp10_prod        | ratings                               |    8176.63 |
| enwp10_prod        | moves                                 |     750.84 |
| enwp10_prod        | global_articles                       |     686.00 |
| enwp10_prod        | selection_data                        |     181.89 |
| enwp10_prod        | cache                                 |      91.86 |
| enwp10_prod        | namespacename                         |      10.55 |
| enwp10_prod        | categories                            |       8.53 |
| enwp10_prod        | reviews                               |       5.03 |
| enwp10_prod        | workingselection                      |       2.52 |
| enwp10_prod        | automatedselection                    |       2.52 |
| enwp10_prod        | builders                              |       2.52 |
| enwp10_prod        | manualselectionlog                    |       0.77 |
| enwp10_prod        | projects                              |       0.34 |
| enwp10_prod        | manualselection                       |       0.30 |
| enwp10_prod        | selections                            |       0.13 |
| enwp10_prod        | zim_files                             |       0.08 |
| enwp10_prod        | users                                 |       0.06 |
| enwp10_prod        | global_rankings                       |       0.03 |
| enwp10_prod        | releases                              |       0.03 |
+--------------------+---------------------------------------+------------+

This was accelerated by the fact that "Articles by Quality" and "Category class" templates were merged, which basically generated a superfluous log entry for every article. However, even if that hadn't happened, the logs table would have eventually run out of space, since it only grows over time.

Since we only ever show logs for the past 7 days, we don't need to hold data older than 7 days. Although #49 would make this easier and bring its own benefits, at the very least we could have a cron job that deletes data older than 7 days, or do it as part of generating the logs.