[mi-dbdata-import] Parallelize import jobs

[SerhatG]

This is done using GNU Parallel.
It has a lot of useful functionality.
Also handles some edge cases and does error handling properly.

It does mean we have to move the add_json_to_collection function to an include file for it to work. That might need some cleaning up.
With my tests an import job uses about 1.25-1.6 CPUs per job, so went for 60% as the default value.

[SerhatG]

Some local numbers:

### Small data job

# Default
Time: 00:51:23
CPU: 125-165%

# 4 jobs
Time: 00:19:57
CPU: 125%-660%

# 60% of the threads (14 jobs on my machine)
Time: 00:11:38
CPU: 125%-2100%

### Bigger data job

# Default
Time: 04:03:23

# 4 jobs
Time: 03:55:06

# 4 jobs - split files
Time: 02:12:23

# 60% of the threads (14 jobs on my machine) - split files
Time: 01:13:15

# 40% of the threads (9 jobs on my machine) - split files
Time: 01:24:57

[SerhatG]

Ah the export part is missing for parse_arguments(). But I'll refrain from fixing it, as this might be one of the things that might need to be tidied up.