Add SparseHist wrapper for large multi-systematic histograms

[bendavid]

Three commits adding a SparseHist wrapper class around scipy
sparse arrays carrying hist axes metadata, plus supporting fixes.

This provides a minimal python representation for sparse boost
histograms in C++ from narf which allows them to be pickled
and/or passed directly to rabbit without creating a dense intermediate.

• Add SparseHist wrapper combining a scipy sparse array with hist
  axes (c833677): stores the dense N-D shape implied by a sequence
  of hist axes in the with-flow layout (axis.extent per axis) and
  provides toarray and to_flat_csr methods that extract either
  the with-flow or no-flow representation. Also supports dict-style
  slicing along axes by regular-bin index for use cases such as
  multi-systematic dispatch in rabbit.

• Use int64 indices in SparseHist.to_flat_csr for large flat
  sizes (256be1f): the CSR returned previously cast indices and
  indptr to int32, which silently overflowed when the flat target
  size exceeded the int32 range. This affected SparseHist instances
  built from large multi-axis inputs (e.g. an
  (eta, phi, pt, mass, corparms) hist with ~108k corparms, where
  the with-flow flat size is ~6.3 billion bins). Now switch to int64
  whenever the target size does not fit in int32.

• Protect against future incompatible change in hist (71f7eb6).

[davidwalter2]

Not sure if this is the idea but if we want to use SparseHist as drop in replacement for a regular Hist object we should give it the same attributes.

Right now "name" and "label" are the obvious ones missing.

There are also small differences e.g. the .shape for SparseHist includes under/overflow while it it not included in the regular Hist.

On the Hist object I can also do things like "h_dense.axes.name" which doesn't work for the SparseHist.

Functions like "fill" or "project" could be set as "NotImplemented" or "NotSupported"

Just for reference this is the full list:

>>> h_sparse.__dir__()
['_axes', '_dense_shape', '_size', '_flat_indices', '_values', '__module__', '__firstlineno__', '__doc__', '_underflow_offset', '__init__', '_from_flat', 'axes', 'shape', 'dtype', 'nnz', 'toarray', 'tocoo', 'to_flat_csr', '__getitem__', '__static_attributes__', '__dict__', '__weakref__', '__new__', '__repr__', '__hash__', '__str__', '__getattribute__', '__setattr__', '__delattr__', '__lt__', '__le__', '__eq__', '__ne__', '__gt__', '__ge__', '__reduce_ex__', '__reduce__', '__getstate__', '__subclasshook__', '__init_subclass__', '__format__', '__sizeof__', '__dir__', '__class__']

>>> h_dense.__dir__()
['_variance_known', 'name', 'label', '__module__', '__firstlineno__', '__static_attributes__', '__orig_bases__', '__weakref__', '__doc__', '__parameters__', '_family', '__slots__', '__init__', '_generate_axes_', '_repr_html_', '_name_to_index', '_to_uhi_', 'from_columns', 'project', 'T', 'fill', 'fill_flattened', 'sort', '_convert_index_wildcards', '_loc_shortcut', '_step_shortcut', '_index_transform', '__getitem__', '__setitem__', 'profile', 'density', 'show', 'plot', 'plot1d', 'plot2d', 'plot2d_full', 'plot_ratio', 'plot_pull', 'plot_pie', 'stack', 'integrate', '__annotations__', '__init_subclass__', '_clone', '_new_hist', '_from_histogram_cpp', '_from_histogram_object', '_import_bh_', '_export_bh_', '__getattr__', '_from_uhi_', 'ndim', 'view', '__array__', '__hash__', '__eq__', '__ne__', '__add__', '__iadd__', '__radd__', '__sub__', '__isub__', '__mul__', '__rmul__', '__truediv__', '__div__', '__idiv__', '__itruediv__', '__imul__', '_compute_inplace_op', '__str__', '_axis', 'storage_type', '_storage_type', '_reduce', '__copy__', '__deepcopy__', '__getstate__', '__setstate__', '__repr__', '_compute_uhi_index', '_compute_commonindex', 'to_numpy', 'copy', 'reset', 'empty', 'sum', 'size', 'shape', '_handle_slice', '_rebin_with_groups', 'kind', 'values', 'variances', 'counts', '_hist', 'axes', '__dict__', '_types', '__class_getitem__', '__new__', '__getattribute__', '__setattr__', '__delattr__', '__lt__', '__le__', '__gt__', '__ge__', '__reduce_ex__', '__reduce__', '__subclasshook__', '__format__', '__sizeof__', '__dir__', '__class__']

>>> h_dense.__dict__
{'_variance_known': True, 'name': None, 'label': None}

>>> h_sparse.__dict__
{'_axes': (Regular(20, -5, 5, name='x'),), '_dense_shape': (22,), '_size': 22, '_flat_indices': array([ 1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 16, 17,
       18, 19, 20]), '_values': array([265., 235., 247., 249., 249., 263., 260., 265., 226., 248., 247.,
       254., 230., 227., 261., 246., 254., 283., 242., 249.])}