FEAT: Pipeline and processor elements

[mberz]

We could consider implementing different SDM variants using pipelines in a similar way to scikit-learn.

This would require that each element has compatible interfaces (for example a process or __call__ function, in scikit learn they are called transform or fit)
In this pipeline logic, the output of one element would always be passed to the respective processing method of the next class.

analysis_step = Analyzer(parametrization_analyzer)
output_analysis = analysis_step(input_signal)

synthesis_step = Synthesizer(parametrization_synthesizer)
output_signal = synthesis_step(output_analysis)

pipeline = SDMPipeline(analysis_step, synthesis_step, postprocessor, post-processor_2)
output_signal = pipeline(input_signal)

pipeline = sdm_classic(parameters)
output = pipeline(input)

[nmkahlen]

I like this! As discussed in the meeting:

What about having different classes for LoudspeakerSynthesizer, BinauralSynthesizer, AmbisonicsSynthesizer?
All would inherit from the base class. BinauralSynthesizer could inherit from LoudspeakerSynthesizer

pipelines would be in the methods module

[nmkahlen]

HO-SIRR and even later down the line: How do they fit in? Is there a SRIR. namespace and a "Running Signals" namespace like SRIR.LoudspeakerSynthesizer