Rc0.4.5 (#144)

* Feat/better defaults (#142)

Updates:
* Modularisation of SwyftModule
* Restructuring of `configure_optimizers` and `configure_callbacks`
* New default is AdamW with early stopping
* Alternatives with AdamW and OneCycleLR and ReduceLROnPlateau exist as well
* Automatic reloading of best model at the end of training is default

* Feat/plot cleanup (#138)

Updates:
* Remove plot2.
* Extended options
* Add `smooth_prior` flag for interpolating likelihoods
* Renaming of plotting routines (plot_posterior, plot_corner, plot_pair)
* Update example notebook

---------

Co-authored-by: NoemiAM <[email protected]>

Release/v0.4.4 (#134)

* Remove gaussian autoregressive models (to come in a later release).
* Update changelog.rst to v0.4.4

Feat/pl2 compatibility (#124)

* Lightning >= 1.5.10

Update README.rst (#121)

Add recent papers using Swyft.

Merge current dev branch (#115)

* Noisy logratios in coverage estimates
	modified:   lightning/core.py

* Black code

* Add example notebook for model comparison

3 Changes to be committed:
	new file:   0I-Model comparison.ipynb

* Use tqdm.auto for better progress bars in notebooks.

	modified:   swyft/lightning/simulator.py

* Enable per-marginal features in default LogRatioEstaimators

	modified:   networks/classifier.py

* Fixes LogRatioEstimator_1dim_Gaussian

 Changes to be committed:
	modified:   lightning/estimators.py

* Black code

* Add `minstd` arguemnt to Gaussian posterior estimator.

 Changes to be committed:
	modified:   estimators.py

* Automatic plotting axes.

 Changes to be committed:
	modified:   plot2.py

* Fix math in docstrings

 Changes to be committed:
	modified:   lightning/estimators.py

* Black code

Fix README.rst for bleach

Update README.rst

Remove deprecation warning, add on_after_load_sample option

	modified:   data.py
	modified:   simulator.py

Remove deprecation warning, add on_after_load_sample option

	modified:   data.py
	modified:   simulator.py

Today's work on empirical mass estimation.

- Plotting routines work now.
- Turns out that mass estimation is not correctly treating HDP intervals