v3.70

Release 3.70 brings a lot of new features:

• blocks interleaving for higher concurrency with simple independent loops
• more control on thread's local stack size
• possibility to pass arguments to parallel loops and pipes as a table of pointers instead of a var_arg list
• possibility to disable block sorting for better cache efficiency at the cost of lower concurrency
• the Hilbert utility has a new GPU-friendly option that sorts entities through their reference and ball degree (vertices)