Remove parallel import code

[craigds]

Description

This code adds complexity and doesn't help since the repo layout
optimisations were introduced in Datasets V3, although
it probably does speed up datasets with string primary keys.

With sequential integer primary keys, it just results in idle processes
since features from a sequence are consistently fed into the same
process, resulting in no efficiency gain.

Since this code is quite complex, it doesn't make sense to keep it

On my laptop, I got these results:

--num-processes=10:

Added 2,328,809 Features to index in 87.5s
Overall rate: 26607 features/s)
Closed in 0s
Joining 10 parallel-imported trees...
Joined trees in 0s

--num-processes=1:

Added 2,328,809 Features to index in 88.4s
Overall rate: 26344 features/s)
Closed in 0s

The idle processes can be seen in Activity Monitor:

Related links:

.

Checklist:

• Have you reviewed your own change?
• Have you included test(s)?
• Have you updated the changelog?