Refl1d

==================
* fixed PolarizedQProbe.calc_Q (should not call PolarizedNeutronProbe._calculate_union)
* added self-extracting self-contained folder install for Windows (see releases page)
* added Error-in-Variables examples (see doc/examples/eiv folder)
* fixed shouldn't show unpolarized data in spin-asymmetry (SA) view (thanks, purnimab)

==================
* preliminary support for theta_offset in PolarizedNeutronProbe
* preserve manual zoom view on R during fits
* fix wx and matplotlib bug (setTitle)
* use numba JIT for reflectometry kernel, convolve, etc (replacing all uses of C/C++ extensions)

* (this results in a pure Python wheel)
* use magnetic_smooth_profile in "record_point", same as is done for non-magnetic profiles
* FunctionalLayer can now accept lists of parameters
* oversampling helper added (currently called refl1d.check_oversampling)
* much faster "Model Uncertainty" view calculation for large profiles
* update to bumps 0.8.1

==================
* numba included in windows zipped executable
* use numba rather than C++ for fast polymer calculations
* wheels now built with stable ABI (one wheel per platform for all Python > 3.2)
* drop py27 support for compiled modules
* tolerance adjusted on polymer_test to account for sensitivity to input parameters
* display of chisq in plot window works again
* fix selection mask for load4 to work on theta and lambda as well as Q and R
* use github actions to build wheels and deploy

==================
* update to bumps 0.8.0, with DREAM outlier removal and convergence tests
* allow garefl to work with DREAM again

==================
* support uniform resolution in addition to gaussian resolution
* implement *oversample()* and *critical_edge()* for *QProbe*
* support variable wavelength probes for materials mixtures
* add irho to the profile contour plots
* save magnetic twist contour in output directory; it is not yet plotted
* simplify realignment usage to 'refl1d align ...'
* improve support for complex amplitude (beam corrections, interpolation)
* restore DREAM fitter efficiency (it should now require fewer burn-in steps)
* restore parallel processing for GUI
* reenable python 2.7 support

==================
* support list values for angle and wavelength in column data format
* use dQ from datafile rather than calculating it from dL and dT
* adjust dQ with sample broadening rather than calculating it from dL and dT
* fix json output so saving the fit doesn't raise an error
* fix functional magnetic profile so [rho, rho] is not treated as rho, theta.
* fix plots to show magnetic angle when not the standard magnetic angle
* fix legends when show magnetic angles on plots
* wxpython is now an optional dependency
* bumps: support *limits=(min, max)* for pm and pmp parameter ranges
* bumps: fix *--entropy* command line option
* bumps: parallel fitting suppressed in GUI for now---need to reuse thread pool

**BREAKING CHANGE**: default thetaM for functional magnetism changed from 0 to 270