contrib.concurrent: Use python default max_workers

[aripollak]

I noticed that contrib.concurrent defines its own default for max_workers, which is the same as CPython's default for ThreadPoolExecutor in CPython 3.8+, but differs from the default for ProcessPoolExecutor:

If max_workers is None or not given, it will default to the number of processors on the machine...If max_workers is None, then the default chosen will be at most 61, even if more processors are available.

This was surprising to me. If the running code is CPU-bound, which is usually why you'd use process_map instead of thread_map, there usually isn't a noticeable speed improvement to spawn more processes than the number of available CPUs. In my case, cpu_count + 4 ended up using more RAM than necessary for maximum speed improvement. I'm using a package that loads an index into memory at 800 MB per process, so that's an unnecessary extra 3.2 GB used by default.

Obsoletes #1530.