Improve DB archiving playbook by using parallel dump

[sbesson]

The DB bump has been quite reliably the bottleneck of the decommissioning process. This tries to enable parallel dump using the directory format

[sbesson]

The execution of the last state of the archive-db.yml playbook (8 parallel threads) on prod94-database completed in 36min split into:

• 19min for the multi-threaded directory dump (162 files, 8.0GB)
• 17min for the gzipped archival (1 file, 7.3GB)

Given the limited gain in terms of disk space, I would consider stopping at the first step and use the directory structure.

[sbesson]

Used for decommissioning prod94