num_features should be extracted from DLC data

[luiztauffer]

as of now we have an input argument num_features that users should pass to the configuration at the start of the pipeline, but if this number does not match the correct number of features from the DLC data, in later steps this might cause errors.

We should:

• retrieve num_features from the input data (e.g. DLC csv)
• output the tuple (index, body part name) at project creation
• allow features to be addressed by name or indexes at functions such as egocentric_alignment

[luiztauffer]

@DrSRMiller @katiekly what's your opinion on this?