Histbook

Fixed bug found in issue 53. `Hist.fraction` has to drop profile axes, but fails when there are no profile axes (it's the trivial case with nothing to drop).

Two-dimensional histogram -> PyROOT.

Thanks, clelange! (PR 51).

Fixes 37 and 42: plotting improvements and a missing `import threading`.

Transfer to Scikit-HEP.

Added

* semantic histogram booking: see comment on 20
* ... which required `Books` to have a hierarchical tree structure
* ... which is aided by wildcard-based selection and viewing (projecting slices to fill independently of the rest)
* normalized histograms in tables (e.g. Pandas) and plotting
* fill from Pandas (`hist.fill(df)`) much like Spark
* JSON-serialization for histograms and books
* histograms have a `filter` as well as a `weight` (although a filter can be emulated by a weight)
* histograms and books can have arbitrary JSON attachments (e.g. fit results)

Several bug-fixes and features:

* Added `PlottingChain.heatmap` for two-dimensional histograms.
* Added a `grid` function for arranging plots in a rectangular grid.
* Fixed `Combination` classes (`overlay`, `beside`, `below`, and now `grid`) to be composable.
* Fixed 24: fill can accept a mix of array and scalar data.
* Fixed 23: named constants (like `pi`, `e`, `inf`, `nan`) are only assumed to be constants if not specified in `fill`.
* Removed dependency on the Meta library.