Sparse grid utilities for IR and Chebyshev representations

Essential contents for a temperature-independent sparse grid

stats: Bose or Fermi statistics.
xgrid: Real space grid points in $x \in (-1, 1)$.
ngrid: Frequency space grid points in Matsubara indices $n$, such that $\omega_n = (2n+\zeta)\pi/\beta$.
wgrid: Frequency space grid points assuming $\beta=1$.
uxl: Transformation matrix from the basis representation l to real space grid points x.
ulx: Inverse of uxl.
u1l_pos: Basis representation for $x=+1$.
u1l_neg: Basis representation for $x=-1$.
uwl: Transformation matrix from the basis representation l to frequency space grid points w.
ulw: Inverse of uwl.
metadata: Group for representation-depend information, for example:
- For Chebyshev: ncoeff for maximum number of coefficients.
- For IR: lambda for dimensionless cutoff $\Lambda$, ncoeff for maximum number of coefficients.

python/sparse_grid/: python code for generating and parsing sparse grid data.
- python/sparse_grid/ir/: IR-specific code.
- python/sparse_grid/chebyshev/: Chebyshev-specific code.
- python/sparse_grid/repn.py: Common module interface for representations.
- python/generate.py: Script for generating HDF5 archives.
c++/: C++ interface for loading and representing sparse grid data.
- green/grids/: Public headers.
data/: Pre-generated data files.
examples: Usage examples.

Python:
- Hiroshi's sparse_ir
- numpy, scipy, h5py, mpmath, ...
C++:
- Green/h5pp: for compatibility with h5py
- Green/ndarrays: for compatibility with numpy.ndarray
- Green/params: for comandline parameters

This work is supported by National Science Foundation under the award CSSI-2310582