Ensure safe_to_tensor moves tensors to the specified device. (#831)

This PR fixes a bug in the `safe_to_tensor` utility: previously it did not move tensors to a new device according to the `device` kwarg which caused issues when there is more than one device available.
The bug went unnoticed for a long while since our circleCI runners do not have GPUs enabled.

Running benchmarks (#812)

* Add script to run benchmarks on slurm

* Increase memory and decrease QOS for benchmark runs.

* Add bash script to run the entire benchmark without any parallelization.

* Add scripts to generate benchmark summary information and computing the probability of improvement.

* Speed up the numpy transform when loading a trajectories from a huggingface dataset.

* Add asserts and explanation for the sample matrix.

* Add script to export sacred runs to csv file.

* Add mean/std/ICM and confidence intervals to the markdown summary script.

* Explain how to run the entire benchmarking suite and how to compare a new algorithm to the benchmark runs.

* Switch from coco's deprecated --side-by-side option to --allow-downgrade.

* Split up the python and openssl/ffmpeg installation to ensure that the step properly fails when the installation of one of the packages fails.

* Add tests for sacred file parsing.

* Add link to the rliable library to the benchmarking README.

* Add no-cover pragma for warnings about incomplete runs.

Documentation nits (#756)

* Fix heading level in tutorial.

* Fix capitalization of algorithm article titles.

Run workflow on version tag push as well

Dropping support for python 3.7 (#505)

* dropped support for 3.7

* added python 3.9 and 3.10 to classifiers

Automate PyPi uploads on release (#489)

* Automate PyPi release

* Use our own PyPi release of sacred (yuck!)

* Bump version number (#490)

* Switch to generating version number from setuptools_scm

* fix typo

* Use v1 release

* Add backwards compatibility logic

* Disable local scheme

* Fetch tags

* use v3 not master

* Bump to 3.8 everywhere

* No cover for clause not expected to hit

* Workaround flaky windows tests

compute_train_stats: Fix logits passed in as proba (#273)

Led to an error when I was training.

Final TF1 release

Clean up for PyPI Upload (#187)

* Update setup.py to prepare for PyPI release

* Remove stray __init__ file (was breaking pytype amongst other problems)

* Seals not benchmark_environments

* Rename benchmark_environments to seals

* Fix SEALS version

* Update configuration guidelines (now horribly out of date)

* Clarify contribution guidelines in README and update code coverage to have stricter check for project/main

* Add PyPi badge in anticipation of our glourious indexed future

* Update Ray to 0.8.x and make dependency optional