Feat: add basic parallel support for listing packages

[GaetanLepage]

When ofborg's evaluation results are not available or when the user specifies --eval local, local evaluation is used to list packages affected by a PR.
This process is quite time-consuming and could be done in parallel when evaluating for multiple systems.

Ultimately, the performance of nix-eval-jobs could be leveraged to perform this task.
Unfortunately, as it is currently, nix-eval-jobs cannot be used as a drop-in replacement in this code base.

As an intermediary situation, I suggest to ship a simple parallel solution.
This approach can be very memory-intensive.
Running it with 4 parallel processes (one for each system) consumes approximately 55GB of RAM.
Nonetheless, it unsurprisingly reduces the time spent on evaluation almost linearly.

I did a naive test running nixpkgs-review --systems all (4 systems) on a PR I had already reviewed (all build artifacts were cached on my system).
The total wall-time was 5:17 with 4 parallel processes and 16:51 with a single process.

An important decision to make is about how to parametrize the number of threads.
#419 introduced --num-procs-eval which is currently used to control how many nix eval commands were ran in parallel.
Although it could be convenient to use this single flag for both processes, the implication in terms of memory usage are significantly different.

How would you see the configuration for this new parallel feature ?

• A new flag ? If so, do we keep the current name ?
• Having two flags ? (Renaming the old one in the meantime ?)

[Mic92]

If you eval the whole of nixpkgs for different systems than this approach is probably even faster because it overlaps less duplicate computation compared to nix-eval-jobs.
Each system is independent from the others.

[GaetanLepage]

If you eval the whole of nixpkgs for different systems than this approach is probably even faster because it overlaps less duplicate computation compared to nix-eval-jobs.
Each system is independent from the others.

Yes, I have been using it lately and it works fine.

However, I am still wondering what we should do w.r.t the user interface.
Should we have two flags to separate the nix-env parallelization and the nix eval one ?

[Mic92]

@mergify rebase

[mergify]

rebase

✅ Branch has been successfully rebased

[Mic92]

@mergify queue

[mergify]

queue

🛑 Command queue cancelled because of a new queue command with different arguments

[GaetanLepage]

Is it stuck ?

[Mic92]

Yeah. Mergify is buggy lately.

[Mic92]

@mergify queue

[mergify]

queue

🟠 Waiting for conditions to match

• -closed [📌 queue requirement]
• -conflict [📌 queue requirement]
• -draft [📌 queue requirement]
• any of: [📌 queue -> configuration change requirements]
  • -mergify-configuration-changed
  • check-success = Configuration changed
• any of: [🔀 queue conditions]
  • all of [📌 queue conditions of queue default]

[GaetanLepage]

Yeah. Mergify is buggy lately.

Ok, thanks for the careful review :)