Future: parallel local and remote build plus ev. heuristics

[gww-parity]

For very small jobs it may be faster to build locally then upload, remotely and download result.

Idea is that it may be started in parallel local and remote build workflows, and whichever finishes faster, that one "wins" and stops the other.

Also there could be heuristics to determine which builds to not even try to build in that way or which not even try to build remotely / locally.