Release version 2.2.0-alpha

What's Changed

* Continuous integration: Migrated to GitHub actions
* Installation: Perl extension disabled by default (--disable-perl argument removed from configure). To enable perl extension, use --enable-perl
* IGRF: Updated to DGRF12/IGRF13. Update extrapolation to use secular variation.
* Lgm_QinDenton: Extra checks in Lgm_QinDenton to reduce unnecessary output and check for inputs with standard QD naming convention ([1min, 5min, hour], in that priority order)
* New functions: Lgm_Terminator, FluxTubeVolume, Lgm_ReplaceSubString2, Lgm_B_FromScatteredData6 added to library
* Experimental Gaussian Process Regression functionality added to library
* Updates to enable second-order evaluation of fist adiabatic invariant
* magstep: Bulirsch-Stoer relative tolerance set to zero by default (only use absolute error)
* API change: Add RelaxTolerance argument to FindShellLine
* API change: Add ErrorStatus to all function variants of ComputeI_FromMltMlat
* TA16: Tsyganenko-Andreeva 2016 model added. Currently requires TA coeficient input files.
* TS04: Trap added for clean return if evaluating deep inside Earth
* JensonCain1960: New internal field model added
* Bug fixes: Correct altitude for lambda integrand in L* calculation, Correct angle in Lgm_EarthEclipse, Correct intersection calculation for Lgm_Ellipse
* Other: Preliminary version of field-aligned (PQB) coordinates, expanded KdTree functionality, updates to various examples, tests updated for expected changes, numerous minor documentation updates, etc.

Full Changelog: v1.5.16...v2.0.0-alpha