makeClusterPSOCK(): Make `workers = availableCores()` the new default

[HenrikBengtsson]

Advantages:

• Easier to write cl <- makeClusterPSOCK() and cl <- makeClusterPSOCK(workers = availableCores())
• Even easier to write cl <- parallel::makeClusterPSOCK() and cl <- parallel::makeClusterPSOCK(workers = parallel::availableCores())

Disadvantages:

• Not obvious from looking at the code alone how many cores are used by default.
• New default will be to use all available cores, which might not be the best design for multi-tenant environments.
• If you only need four workers, you might not be bothered to do cl <- makeClusterPSOCK(workers = 4) when all you have to do is write cl <- makeClusterPSOCK(), especially if the latter works just fine on your local 4 or 8 core machine. Then you run the same on a 96 core machine.

[HenrikBengtsson]

An even better default might be makeClusterPSOCK(workers = availableWorkers()), which falls back to rep("localhost", times = availableCores()) if no other information is available.

Having said this, I'm gonna take the conservative approach and not set a default. It's better/safer to always specify workers explicitly.