synchronize with pocketfft C++

See merge request !43

two bug fixes (should not have caused any problems yet)

See merge request !42

Handle keyword args in good_size

See merge request !41

Threadpool fixes

Closes #14

See merge request !40

Avoid putting threads to sleep if there are any unscheduled tasks in flight.

Fix arm neon detection

See merge request !39

Fix feature detection macro for hardware_destructive_interference_size

See merge request !38

Threadpool: Assign work preferentially to threads in creation order

Closes #13

See merge request !37

Enable vectorization on arm when compiled with -mfpu=neon

See merge request !36

address more problems with clang (and follow its optimization hints)

See merge request !35

try to fix clang version detection

See merge request !34

move global variables into static functions

See merge request !33

Use qualified std::thread to avoid name collisions on AIX

See merge request !31

Export good_size using raw C-API

See merge request !32