Pennylane-rigetti

* Added parametric compilation for `forest.qvm` and `forest.qpu`, leading to significant speed increases when performing optimization

* The plugin has been ported to use the new [`QubitDevice`](https://pennylane.readthedocs.io/en/stable/code/api/pennylane.QubitDevice.html) API for plugins, resulting in a cleaner interface with built-in support for multi-qubit observables.

* Bug fixes to allow the plugin to work with pyQuil >= 2.16

* Operator estimation on `forest.qpu` now works with both 1 and 2 qubit gates

* Adds a converter so that pyQuil and Quil programs can be loaded as differentiable PennyLane templates

---

* Added ability to return multi-qubit expectation values to all devices

* Added variance support

* Replaced dense matrix expansion with a tensordot implementation

* Added support for `samples`

* Readout Error Mitigation using Operator Estimation

* Updates the plugin to work with PyQuil 2.13 and PennyLane 0.6

---

New features

* Updated to support PyQuil v2.9.0 and PennyLane v0.4.0. Note that PennyLane-Forest now requires PennyLane>=0.4

* The observable `qml.Hermitian(A, wires)` can now be measured on an arbitrary number of wires. That is, you can now pass a matrix A of size `[2**N, 2**N]` which acts on `N` qubits (i.e., `len(wires)==N`).

* Adds support for the `qml.var()` measurement return type provided by PennyLane v0.4.

---

* Using any of the available lattices at any given time, instead of a specifically named one

* updated minimum pyquil version requirement to 2.7

* added support for pyQVM

* removed 2.4 changes

* Updating compiler urls

* Replacing old lattice with new

* pyqvm working without the compiler

* Updating compiler urls

* Replacing old lattice with new

* added build and documentation badges

* Remove the LocalQVMCompiler from the tests