Rt1

new
- HDF5 containers can now be used instead of pickle-dumps to store `Fits` and `MultiFits` objects
- `RTprocess` is now directly appending results to an HDF5-container during parallel processing
- colored log-messages can be used (requires `colorama` package)
- there's now a proper container for `defdict` to simplify get/set of variable definitions
- indicate fit-results in `fit.plot.analyzemodel`

changes
- `processfunc` is now private (e.g. `_processfunc`) -> use `run_processing` instead!

fixes
- several bugfixes along the way

new
- there's now a quick and easy data-export function `RTprocess.export_data()`
- extract parameters and performance-metrics from a folder of dump-files
- there's now an `RTresults.dump_fits` - generator that directly returns Fits-objects
- `RTprocess` object can now be initialized without a config-file
- option to pass a list of dump-files to `rtprocess.run_finalout()` added


fixes
- speedup for indexing of folders with large amounts of dump-files
- some minor logging updates

new
- add possibility to provide numeric V & SRF functions for faster evaluation if no interaction-term is calculated
- there is now a property `fit.config_name` that indicates the config-name for fits that belong to a `MultiFits` object
- datasets provided to `rtfits.Fits` are now checked for sorting (and automatically sorted if necessary)

fixes
- make `ncoefs` a propery for "Rayleigh" and "Isotropic" phase-functions
- improved logging messages
- don't log each successful progress step, only errors and warnings
- don't create a single file for each error (log traceback to log-file directly)
- keep minimal info from `fit.fit_output` when dumping fits to disc
- improvements on caching and pre-computation steps
- pre-evaluate fn-coefficients during finalout-generation
- cache lambda-functions for `V.p` and `SRF.brdf`
- fix sorting of fractional contributions in `rtplots.printsig0analysis()`
- some updates on `rtplots.polarplot()` for symbolic parameter specification
- make sure copying of config-files and modules is only attempted from the main process

new
- add possibility to provide numeric V & SRF functions for faster evaluation if no interaction-term is calculated
- there is now a property `fit.config_name` that indicates the config-name for fits that belong to a `MultiFits` object
- datasets provided to `rtfits.Fits` are now checked for sorting (and automatically sorted if necessary)

fixes
- make `ncoefs` a propery for "Rayleigh" and "Isotropic" phase-functions
- improved logging messages
- don't log each successful progress step, only errors and warnings
- don't create a single file for each error (log traceback to log-file directly)
- keep minimal info from `fit.fit_output` when dumping fits to disc
- improvements on caching and pre-computation steps
- pre-evaluate fn-coefficients during finalout-generation
- cache lambda-functions for `V.p` and `SRF.brdf`
- fix sorting of fractional contributions in `rtplots.printsig0analysis()`
- some updates on `rtplots.polarplot()` for symbolic parameter specification
- make sure copying of config-files and modules is only attempted from the main process

new
- add possibility to call finalizer after `rt1.rtprocess.run_processing()` finished
- to *"clean-up"* after processing is done (or the process is terminated)
- add option to save dumps in `rtprocess.run_processing()` even if `postprocess` is not called via
`rt1.rtprocess.run_processing(..., dump_fit=True)`

**Multiconfig Fits**
- add possibility to specify multiple configs in a single .ini file
- processing of multiple configs involves only a single call to `read()`!
- you can specify as many configs as you like in a single `.ini` file!
- to specify multiconfigs, add the config-specific changes to sections starting with `[config <NAME> <SECTION>]`
for example:

ini
[defdict]
--- the default defdict ---

[config NAME defdict]
--- all specifications that differ from the default defdict ---


- Multiconfig fits will be handled with a `rt1.rtfits.MultiFits` object
- use `rt1.rtfits.MultiFits.accessor` to access `Fits` propeties of all configs
- all processed configurations can be saved in a single `.dump`-file!
- this avoids duplicating data that is shared between the configs


fixes
- fix issue in the default *"domainfuncs"* of `rt1.general_functions.scale()`
- make sure `rt1.rtfits.reinit_object()` only copies relevant properties
- speed up `rtprocess.RTresults` initialization for folders with many files

new
- logging implemented
- incorporate basic NetCDF export functionality via `xarray` (`rt1.processing_config.postprocess_xarray()`)
- new `rtmetrics` class to provide easy access to common metrics calculation
- add option to disable postprocessing va `RTprocess.run_processing(postprocess=False)`
- add option to run processing and NetCDF export in 2 separate runs
- e.g. via `run_processing()` and `run_finalout()`
- add `dump_fit` option to `run_processing()` so fits are dumped even if `postprocess=False`

changes
- `copy` is now an argument of the `__init__` of `RTprocess` instead of `RTprocess.run_processing()`

unittests
- switch tests to github actions
- use conda `.yml` file for requirements
- switch coverage reports to codecov
- reactivate stickler & flake8
- use github actions for PyPi upload

fixes
- fix processing and logging with `ncpu=1`
- enure that the `processing_config` class and additional modules are loaded from the correct folder
- fix future warnings in `sympy` and `matplotlib`
- make `RTresults.load_fit()` find fits faster
- allow passing integers to load the nth result
- fix sorting in `printsig0analysis()`
- fix parsing error in `param_dyn_dict`
- use pickleable decoration of initializer function
- proper detection of submodules in `setup.py`
- incorporate pending changes in `scipy > v1.6.3` (e.g. adjust call signature of `sympy.Lambdify`)