Elfi

------------------
- Update available methods list
- Pin `numpy <= 1.24.0` due to `GPy`
- Update RTD configuration

------------------
- Fix broken 0.8.5 by adding missing `__init__.py` to elfi/methods/bsl/

------------------
- Fix the option to continue inference in model-based inference
- Move classifiers
- Fix readthedocs configuration
- Update penalty to shrinkage parameter conversion in synthetic likelihood calculation
- Update BSL pre sample methods
- Update BslSample
- Move ROMC tests
- Update to numpy 1.24
- Update readthedocs configuration
- Restrict numpy < 1.24 until codebase has been updated
- Update documentation-related files: docs, conf.py and requirements-dev.txt
- Update PULL_REQUEST_TEMPLATE.md
- Drop tests for py36 and add tests for py39 and py310
- Fix couple of minor bugs in `ar1`-model
- Update parent class in BOLFIRE
- Fix semiparametric synthetic likelihood with glasso/warton and add tests
- Fix plot marginals and remove plot summaries
- Fix stochastic volatility example
- Improve batch simulations in toad example
- Remove synthetic likelihood node and update BSL data collection
- Fix M/G/1 example
- Fix scratch assay example
- Add scratch assay example
- Add GP classifier for ratio estimation
- Fix multidimensional indexing in daycare example
- Add BSL method

------------------
- Modify Lotka-Volterra model's priors as many methods do not support discrete random variables.
- Fix acquisition index in state plot
- Reformat `summary()` for `Sample(ParameterInferenceResult)`
- Fix linting in `arch.py`
- Add summary statistics to Lotka-Volterra model
- Add boolean `observation_noise` option to `lotka_volterra`
- Add parameter names as an optional input in model prior and fix the parameter order in priors used in BOLFI and BOLFIRE
- Add feature names as an optional input and make training data size a required input in BOLFIRE
- Fix the observed property in simulator nodes
- Fix default outputs in generate
- Add docstring description to ARCH-model
- Make MAP estimates calculation in BOLFIRE optional and based on log-posterior
- Use batch system to run simulations in BOLFIRE
- Use `target_model.parameter_names` from instead of `model.parameter_names` in `BOLFIRE`
- Extract BO results using `target_model.parameter_names` from instead of `model.parameter_names`
- Update tox.ini
- Add option to use additive acquisition cost in LCBSC
- Change sigma_proposals-input in metropolis from list to dict
- Fix is_array in utils
- Fix acq_noise_var-bug in acquisition.py. Influenced BOLFI.

------------------
- Add a new inference method: BOLFIRE
- Fix the hessian approximation, visualizations and the line search algorithm in ROMC
- Add tests for all ROMC parts

------------------
- Relax tightly pinned dependency on a version of dask[distributed]
- Change lotka-volterra priors to follow the given reference
- Fix README.md badges
- Fix a few small issues with CONTRIBUTING.rst
- Add Github Actions based CI workflow
- Add the skeleton of TestBench-functionality for comparing methods
- Fix a bug of plot_traces() not working if there is only 1 chain
- Fix histograms in pair_plot diagonals and improve visual outlook
- Improve axes creation and visual outlook
- Fix a bug where precomputed evidence size was not taken into account when reporting BOLFI-results
- Fix a bug where observable nodes were not colored gray when using `elfi.draw`
- Add `plot_predicted_node_pairs` in visualization.py.