Add Model Predictive Path Integral (MPPI) control as an optimizer

[brunzema]

Added Model Predictive Path Integral (MPPI) control as an optimizer. I use MPPI for gradient free MPC in one of my projects and it made sense to me to include it here. In my experience, it tends to perform better than CEM in practice. The implementation follows Williams et. al 2017 (but without the receding horizon). To summarize, this pull request

• adds MPPI as an optimizer
• adds a test for MPPI to tests/optimizers_test.py
• adds a benchmark for MPPI in benchmarks.

Some of the fixes in #12 and #13, namely the frozendict as a default for hyperparameters, are also included which was necessary to make things run.

This is my first pull request to a bigger project. Therefore, please tell me if changes are required! :)

[google-cla]

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

View this failed invocation of the CLA check for more information.

For the most up to date status, view the checks section at the bottom of the pull request.