All of the benefits of parking_lot::Mutex over the std one are listed in its
readme, but the ones that apply
to us are copied here:

• Mutex and Once only require 1 byte of storage space, while Condvar and RwLock only require 1 word of storage space. On the other hand the standard library primitives require a dynamically allocated Box to hold OS-specific synchronization primitives. The small size of Mutex in particular encourages the use of fine-grained locks to increase parallelism.
  • less allocation is always good for performance, especially when all PyObjects are allocated and most require Mutexes
• Uncontended lock acquisition and release is done through fast inline paths which only require a single atomic operation.
  • good, since a lot of Python code will only be run single-threaded
• RwLock takes advantage of hardware lock elision on processors that support it, which can lead to huge performance wins with many readers.
  • requires nightly, but we could either build release binaries on nightly or wait until this is possible on stable
• RwLock uses a task-fair locking policy, which avoids reader and writer starvation, whereas the standard library version makes no guarantees.
  • not sure if this is super necessary but it would definitely benefit the GIL-less experiment we're doing here