Camb

* Support for calculating only [time source transfer functions](https://camb.readthedocs.io/en/latest/results.html#camb.results.CAMBdata.calc_transfers), and then power spectra later. This allows calculating non-linear lensing results for different initial power spectrum and non-linear model parameters without recomputing the time transfer functions.
* Fix for Python raw lensing cross CMB output
* Allow k_per_logint to give denser k sampling even when used with non-linear lensing

* Support for wider range of input parameter values and more error propagation to Python
* Added k_hunit option to get_[non]linear_matter_power_spectrum
* get_matter_power_interpolator will now do log(-pk) interpolation for negative spectra
* Fix to assign Neff when omnuh2=0 (thanks tilmantroester)
* Doc improvements for matter power variables, conda install, mathutils
* Fix for 66, 67
* Tweak to meta class to allow introspection of CAMB's Python classes to work
* Setting thetastar does not initialize reionization model or pk/windows (optimization, avoid errors)

* Fix for get_matter_power_interpolator extrapolation 61
* Fix for zstar calculation with tensors 62
* Tweaks for non-linear warnings and error handling

* Avoid re-computations when getting non-linear power spectrum
* Propagate errors in HMCode to python exceptions and small tweaks
* Bug fix for very small neutrino masses and debug compilation
* Fix for setting lmax in Python for some non-flat models

* HM_code updates for numerical stability, speed, fix for linear massive neutrino growth (thanks A. Mead; can change results and significantly faster)
* Fix for Mac ifort compatibility (thanks C. Kreisch)

* fix for z_drag with high matter densities and for P(k) without reionization
* fix memory leak freeing python result objects
* forutils compatibility update
* Exposes HM code baryon parameters to python