GitLab

Add driver and support for testing across modules. Currently,
no real checks are made that something is actually a regression.

These tests can be used for one check realistic scenarios, that
solver reach convergence (or at least some norms get minimized).
Second some performance data is gathered, which should make this useful
to decide if a change actually speeds up anything.

Basic integration test setup to run full reconstructions. This is a
first step into the direction of having more comprehensive integration
tests, which are machine tested and run.

Old versions had virtual functions an underscore prefix, but now they've
got "Impl"-suffixes. In some comments, this change was not yet done,
this commit changes these places.

Instead of using extra bool flags to indicate whether or not
`LinearResidual` has an operator or data vector, use the type system to
better express intent.

A non existing operator is represented with an null pointer. And the
data vector is wrapped in an optional, which first is clearer in intent
but also removed the need for an unnecessary heap allocation.

To remove naked new in DataHandlerGPU and DataHandlerMapGPU, Badges are
introduced. This way the constructor of DataHandlerMapGPU is private,
but `std::make_unique<DataHandlerGPU>` can still use it through badges.
Without badges `std::make_unique` couldn't find the private constructor.

Write 2D containers to the PGM file format. This can be used to quickly
output and debug images.

add new constructor for ISTA/FISTA, remove dead code, add override, improve lasso tests, convert Problem to LASSOProblem, pass by value

Fixed weird formatting, still not really tests but should work

Faster generation time for already preset dependencies

Adding a function taking `ostream`, which the logger will write out to.
Our current main usage is testing. Checking if messages are print
correctly or not.

Restructured `Logger` a bit to hold a static pointer to a
`spdlog::sinks::dist_sink_st`, instead of multiple sinks per logger.
All logger instances share this sink.

Remove constructors for VolumeDescriptor, that have a initiazlier list and a eigen vector,
leave only the ones with only initializer list, or eigen vectors to init the descriptor

Bump CMake version to 3.14 (needed by CPM) and enforce it; Remove cmake_minimum_required from all but root CMakeLists.txt

Improve tests for SoftThresholding, manually generate proximity_operators in hints, throw exceptions for invalid states in SoftThresholding

- Update CI stage to reflect update
- Update ci_scripts to reflect update
- Reformate code with clang-format-10
- Update contribution guide

The width and height of a detector in the CircleTrajectorGenerator is based on the size of the volume. To avoid
undersampling of the corners of the volume, the size of the detector is now the size of the volume
scaled by sqrt(2). This is closer to how matlab behaves

restructured Docerfiles

- Bump CMake to 3.18.2
- Bump oneDNN to 1.6.2
- Bump to Ubuntu 20.04
- Seperate images for clang-format, clang-tidy, coverage, Cuda build
  with clang, gcc and clang with python bindings
- Dockerfiles for clang and gcc rely on the package manager for the
  compilers
- Updated test scripts, they test more and should error our earlier
- Nightly docker job created
- Docker build script updated
- Include CI job, to build docker images on changes

Coverage includes build folders of python bindings, which pull down the coverage unnecessarly, therefore they are not generated for the coverage CI job

Turn python bindings off for checking with address sanitizer, as they create problems either with linking, or many false positives (see #83 for more info)