Add ONNX support and Benchmarks

What's Changed

• Preserve symlinks in CI by @tkoskela in #359
• Make sure ORT lib location is known to cmake by @20DM in #358
• Fix bugs in examples by @tkoskela in #357
• ONNX Testing by @mmcleod89 in #343
• Remove spdlog reference from README by @20DM in #361
• Move models out of data and install them when onnxrt is on by @tkoskela in #369
• Enable ONNXrt in the CI by @20DM in #368
• Remove the notinstalled data namespace since data is being installed by @tkoskela in #370
• Use non-greek sopt interface by @mmcleod89 in #371
• add Homebrew lib dir to lib path by @20DM in #372
• Update benchmarks by @tkoskela in #367
• Update changed benchmark file name in CmakeLists by @tkoskela in #374
• Rename nu and gamma in padmm examples by @tkoskela in #377

Full Changelog: v4.2.0...v5.0.0

Update build system and tests

• #346 Should make tests and benchmarks compatible with Spack
  • Test data is installed to install directory
  • Tests, examples and benchmarks read data from install directory
• #350 Update CI
• #344
  • Remove Conan support
  • Migrate cppflow interface to ONNXrt

Updated build and logging

Corresponds with SOPT v4.1.0

Initial Tensorflow Model Compatible Release

• Add compatibility for tensorflow model in the forward-backward algorithm.
• Shift to Conan based build system for dependencies.

CMake and DOI update

• The main update in this release is the CMake cod and DOI yaml file.
• Also caching of MPI has been added to travis so that MPI tests pass when the should
• A few small test issues around the power method have been fixed

MPI and Widefield Imaging Release

• Changed interface to use a config file that is a .yaml format.
• Added support for reading measurements from the .uvfits format
• Added support MPI ADMM
• Added the Forward Backward algorithm
• Added GPU option when using arrayfire
• Added ability to simulate observation and perform reconstruction
• Added ability to simulate observation from an input uvw coverage
• Added support for the fast and accurate w projection kernel integration
• Added support for MPI w-stacking algorithm
• Added support to combine MPI w-stacking and w-projeciton
• Added support for distributing measurements across MPI nodes
• Added k-means clustering algorithm to distribute measurements into w-stacks
• Added benchmarks of MPI and serial operators

First release

Contains C and python wrappers for purify.