Rt1

changes
- the reader-function in `rt1.rtprocess` is now called without dict-unpacking!!
(e.g. via `reader(reader_args)` instead of `reader(**reader_args)`)
**! this might require modifications in existing scripts !**

fixes
- correct parsing of `None` in `rt1.rtparse`
- multiprocessing pool initializer also executed in single core processing (for debug)
- `rt1.rtprocess` now uses `warnings.warn` instead of `print` for general messages

new
- convenience-method `interpolate_to_index()` added to `rt1.genreal_functions`

changes
- `rtfits.preprocess()` is now called prior to parallel-processing
- a check if files already exist should now be done within the reader-function!
- copying of modules is shifted from the `rtparse` to `rtprocess` to avoid copying files when config is read
- the configparser now has methods `get_module('modulename')` as well as `get_all_modules()` to programmatically import modules
- all parallel-processing functions have been shifted to the dedicated `rtprocess` module

fixes
- corrections to the treatment of `integer` parameter-dynamics specified in `fit.defdict`
- the cache is now only cleared at the beginning of a `fit.performfit()` call
- correct final message in `processfunc()` progress-bar

new
- a dedicated class for parallel processing (`rt1.rtprocess`) has been introduced
- `rt1.rtprocess.RTprocess` can be used to setup and run parallel jobs
- `rt1.rtprocess.RTresults` can be used to provide easy-access to processing-results

- some updates on rtplots.printsig0analysis()
- now one can select additional timeseries from both fit.res_df and fit.dataset
- ... and the styles can be set using the printparamstyles - dict

changes
- manual parameter dynamics are no longer automatically used if a column `'param_dyn'` is provided
... to use manual dynamics, you now must use one of the following:
- explicitly specify parameter-dynamics as `'manual'` in `fit.defdict`
- to combine a pandas datetime-offset `freq` with a manual-dyn use `'freq + manual'`

new
- the `model_definition` text has been updated to include more details
- the property `fit._model_definition` is added to allow accessing the model_definition string
(useful for example to write it to a file)
- `RT1_configparser` now provides a function `.get_module(modulename)` to import a specific module from a file

fixes
- fix import-location for copied modules in `rtparse` (now modules are imported from the copy-path)

new
- define parameter-grouping with respect to number of observations within a group

fixes
- fix dynamic startvalue-assignment
- better implementation of rectangularize function

- added possibility to provide arbitrary start-values via the dataset
- dataset -columns "parameter_start" are now used to assign start-values for the parameters

- there is now a `load()` function in `rt1.rtfits` to load fit-configurations dumped via `fit.dump()`

fixes
- cache mask during fit
- interpolation now works also with unsorted groupings
-> now it's possible to fit interpolated seasonalities etc.
- ... some speed-ups

- add possibility for parsing (dtype-converted) lists in `rtparse.get_process_specs()`

fixes
- fix of issue with inner-derivatives in case unequal groupings are used within sub-parameters
- speedup in the initialization of the jac-assign rules
- fix issue in `rtplots.analyzemodel` with `int_Q = True`

changes
- change of folder-structure used in default parallel-processing funcs
- put all files in sub-folders of dumpfolder (`cfg`, `results`, `dumps`)
- change of datetime format-specifications in .ini files
- "fmt=" is now used as separator instead of ","