Openfermion

Starting with this version, Python 2 is no longer supported. Other notable changes and additions:
- Functions to partition qubits into subsets for efficient measurement of Pauli strings: `openfermion.binary_partition_iterator`, `openfermion.partition_iterator`, `openfermion.pauli_string_iterator`.
- Added `structure` argument to `openfermion.geometry_from_pubchem`. It is recommended to set this to '3d' to ensure correct structures are returned (see https://github.com/quantumlib/OpenFermion/pull/544).
- `openfermion.get_number_preserving_sparse_operator`.
- `openfermion.prony`.

Notable changes and additions:
- Adds a new class, MajoranaOperator, to represent operators in the Majorana basis. Support for this operator added to `openfermion.jordan_wigner` and `openfermion.bravyi_kitaev`.
- Speeds up construction of jellium Hamiltonians by utilizing translation invariance
- `openfermion.wedge`
- Adds some utility methods to InteractionOperator
- Utilities for tapering off qubits using stabilizer conditions: `openfermion.utils.taper_off_qubits`, `openfermion.utils.reduce_number_of_terms`
- Fixes a bug that caused incompatibility with scipy 1.3.0

Notable changes and additions:
- The `diagonalizing_bogoliubov_transform` method of `openfermion.QuadraticHamiltonian` now returns a tuple of three objects: the orbital energies, the transformation matrix, and a constant energy offset
- `openfermion.FermiHubbardModel`
- `openfermion.IsingOperator`
- `openfermion.group_into_tensor_product_basis_sets`
- Improves low-rank decomposition to work in spatial orbital basis rather than spin-orbital basis
- SymbolicOperator is now an abstract base class
- Specialized routines to compute commutators of Diagonal coulomb operators faster
- Updated Docker image to include [OpenFermion-Cirq](https://github.com/quantumlib/OpenFermion-Cirq)
- Default absolute precision for floating point operations changed from 1e-12 to 1e-8

Eighth production release. Minor release necessitated by depending libraries.
* Most importantly, new logo.
* Functionality for performing low rank decompositions of two-body operators.
* Small additions of helper functions for more flexible testing.

Seventh production release. Many new features introduced including:

* Method to obtain circuit that diagonalizes a QuadraticHamiltonian.
* Generalized equality and arithmetic operations for PolynomialTensor.
* Fixed minor bug in is_hermitian function.
* New DiagonalHamiltonian class for interacting fermion systems with diagonal two-body term.
* Human-readable IO for SymbolicOperator.
* Functions to map particle RDMs to hole RDMs and vice versa.
* Fixed bug in plane wave / dual Hamiltonian generation which occurred for even grids.
* Code to preprocess coefficients for LCU algorithm introduced in arXiv:1805.03662.
* LinearOperator class for efficient numerical diagonalization of QubitOperators.
* Routines to use Davidson routine for obtaining ground states using LinearOperator.
* Functions to automatically obtain MolecularData geometries from PubChem database.
* New BosonOperator class.
* Generation of Bose-Hubbard Hamiltonians.
* Functions to automatically remove redundant qubits post-BK transformation.
* Added functionality to generate plane wave / dual basis Hamiltonians with reduced periodicity.
* Links to Xanadu Strawberry Fields "OpenBoson" library.

Unfortunately a change between v0.4 and v0.5 introduced a circular dependency in the setup script that caused an install problem with the v0.5 release for some environments. The v0.6 release corrects that issue and also reconfigures our continuous integration testing so that problems like this will be more easily detected in the future.