Float 32 output, HDF5 overhaul, static_grav and other minor fixes.

[alwinm]

• added make.type.rot_proj and make.type.static_grav for ease of debugging both flags

• added float32_sound_wave.txt as an example of using 32-bit float output

• removed unused minof3 and testfunction from global_cuda.cu

• rename gravity_cuda.h to static_grav.h and put inline device functions into header file to be used in hydro_cuda.cu

• GPU-accelerate 3-D hdf5 output (1-D or 2-D not really worth the effort imo)

• hdf5 output refactored

  • writing a double buffer uses the double precision HDF5 macros and writing a float buffer uses the single precision HDF5 macros using function overloading
  • repacking with/without type conversion selected through function overloading (Real -> Real vs. double -> float)
  • Some fields left un-refactored in situations where I didn't know how to test them.

• Float 32 output

  • No compiler flag. Enable with input parameter n_out_float32 to set output cadence
  • Enable individual fields with out_float32_[fieldname]
  • GPU-accelerated conversion + repacking
  • 3-D only, won't crash code in 2-D or 1-D.