Matlab tests, enhance build

Improve build for Python and Matlab

fix/add Matlab tests and examples

significant calculation bugfix

• require geomagindices >= 1.4.0 for improved f107s and Aps computation that drops NaNs
• this improved algorithm makes MSISe00 here nearly exactly match CCMC
• regenerated test data
  added more error checks

better UX, type hinting

• redo type annotation for Numpy 1.20 and Python 3.10 style (compatible with Python >= 3.7)
• give better error if CMake not found pip install cmake

reliability via Python >= 3.7, new geomagindices data source

• SWPC obsolesced their prior data source, this used geomagindices >= 1.3.0 to use the new recent data source to provide Ap and f10.7 to MSIS
• use Python >= 3.7 for reliable, cross platform and Python VM agnostic access to filesystem artifacts

src/ layout, entry_points, add units

• use src/ layout for modernization
• entry_points for robustness
• document units
• add diagnostic info printout if runtime fails
• automatically run test for build on run, to catch compiler / runtime library issues at first install

ci refactor, use cmake as default

v1.7.2

ci: per-os workflows

add ninja to python prereqs

v1.7.1

add ninja prereqs

robustness improvements

• default to Meson build system for "build on run" for both Python and Matlab
• handle self-test tolerance more reasonably. Be more specific about Fortran format output across compilers, to avoid bugs
• check the returned data from Fortran for gross errors
• move to GitHub Actions CI

enhance packaging

use package_data via MANIFEST.in to handle the Fortran source code for any installation method, in particular from PyPi tar.gz.

Auto-build for Python and Matlab, increase test coverage

• breakup tests for clarity
• autobuild msis for Python and Matlab
• switch to Azure CI entirely. May go to GitHub Actions in future.