WIP: Incorporate quaddtype

[dlakaplan]

This is an attempt to allow native operations on a wider range of architecture using https://github.com/numpy/numpy-quaddtype to handle the extended precision (see #1944). Most things work, but a few changes still need to be made on the numpy side:

• np.average(), np.histogram(), and np.dot(): see [BUG] np.average with weights fails numpy/numpy-quaddtype#106 FEAT: Missing vecdot, matvec and vecmat ufunc implementation numpy/numpy-quaddtype#71.
• Multiprocessing with gridutils breaks, probably related to pickling (Errors using with concurrent.futures numpy/numpy-quaddtype#107). That may be harder to fix.

[dlakaplan]

Test setup procedure:

conda create -n pintdev_arm --platform=osx-arm64 astropy matplotlib ipython
conda activate pintdev_arm      
pip install numpy_quaddtype
pip install -r requirements.txt         
pip install -r requirements_dev.txt     
pip install -e .

[dlakaplan]

Failing tests (locally):

• tests/test_event_optimize.py - needs np.histogram
• tests/test_fermiphase.py - needs np.histogram
• tests/test_gls_fitter.py - needs np.dot
• tests/test_grid.py
• tests/test_model_ifunc.py
• tests/test_noise_models.py - needs np.dot
• tests/test_noisefit.py - needs np.dot
• tests/test_photonphase.py - needs np.histogram
• tests/test_random_models.py - needs np.dot
• tests/test_reduced_precision.py
• tests/test_residuals.py - needs np.dot
• tests/test_utils.py - needs np.dot