Flavio

The major new physics feature in this release is the addition of the branching ratio of the rare _inclusive_ B→X_ql⁺l^- (l=e or μ) decay. Many thanks to Tobias Huber for providing analytical expressions and making numerical cross-checks that made the implementation of (most of) the corrections in [arXiv:1503.04849](https://arxiv.org/abs/1503.04849) possible. Thanks also to the Munich Institute for Astro- and Particle Physics (MIAPP) for hospitality during this effort.

Another new feature is more cosmetic: metadata have been added to all observables, making it possible to split the [list of observables](https://flav-io.github.io/docs/observables.html) into categories rather than one huge list.

Finally, thanks to Christoph Niehoff for improvements and cross-checks in B_q→l⁺l^-.

This release does not bring any new features but various speed optimisations, most significantly:
- Observables in B→Vll at very low q², that used to be slow to compute, are now a factor of 50 (fifty) faster! (318944ed1762b1102a0c81091ae99196d0faca3c)
- Generating random parameter sets is now 4 times faster, which also speeds up the generation of pseudo measurements for [`FastFit`](https://flav-io.github.io/docs/fits.html) (0758c678b399d8b2cb493e952d01d382fa82b7fa)

News in this release:
- New observable: effective lifetime in B_s→ll added. Thanks to ChristophNiehoff for this addition!
- Improvements in the plotting function `flavio.plots.band_plot`
- It is now possible to save the data corresponding to any individual likelihood plot, which allows to easily change or rearrange plots later without having to recompute the likelihood. Thanks to Albert Puig for this addition!
- It is now possible to interpolate the likelihood in between computed points to obtain smoother contours without computing on a finer grid. For the time being, this feature must be activated manually by setting `interpolation_factor` to an integer greater than 1.
- Typos in measurements fixed and measurements of Λ_b→Λμμ added

This point release features a few minor improvements:
- Two typos in measurements fixed (thanks to S. Reichert for spotting one of them)
- Improvement in plot functions: `band_plot` now plots multiple confidence levels and supports legends out of the box
- Stability improvements for `FastFit`

The focus in this release is on experimental measurements and treatment of likelihoods:
- A large number of measurements by LHCb have been added - thanks to Stefanie Reichert for that!
- While by default only the most recent measurements of an observable by a given experiment are included in the code, data files for older measurements can be found in a [separate repository](https://github.com/flav-io/measurements) now.
- A new function to read measurements from a URL can be useful for this purpose: for instance

python
flavio.measurements.read_url(
'https://github.com/flav-io/measurements/blob/master/2011-lhcb-bksmumu.yaml')


All existing mesurements can be cleared with the new method `flavio.Measurement.clear_all()`.
- The combination of univariate likelihoods has been completely rewritten. For parameters or measurements with several uncertainties, these are now combined by computing the convolution of the PDFs. [Here is an example](https://gist.github.com/DavidMStraub/68be4832f1e2e52cb04aa9ca06e4c980) for a symmetric and an asymmetric uncertainty. Thanks to Jens Jasche for discussions.
- For measurements, instead of central values and uncertainties, now also upper limits can be specified. In the YAML data file, they are written simply as a string of the form `< 1.5e-8 95% CL` - this works for arbitrary confidence levels.

Finally, there are two new observables: the direct CP asymmetries in B→K*ll and B→Kll.

This release adds observables in three more decays:
- Branching ratios of D→lν and D_s→lν decays (where l=e or μ) that are used to measure the CKM elements V_cs and V_cd
- Branching ratio and angular observables of Λ_b→Λll (where l=e or μ) that is based on the b→sll transition and is closely related to B→(K, K*)ll.

The Λ_b decay is based on the lattice form factors of [arXiv:1602.01399](https://arxiv.org/abs/1602.01399) and the angular distribution derived in [arXiv:1410.2115](https://arxiv.org/abs/1410.2115). Special thanks to Danny van Dyk for discussions.

On an unrelated note, there is now a [repository with example Jupyter notebooks](https://github.com/flav-io/flavio-examples) that will showcase some of the features in the future (and contributions are of course welcome).