cov_model and known_cov now parsed

[adeliegorce]

Resolves #327

• Added cos_model and known_cov arguments to functions related to data covariance in order to properly parsed the potential user-input covariance matrix/dictionary. These include: iC, cov_q_hat, q_hat, get_G, get_H, get_unnormed_V, get_analytic_covariance, scalar_delay_adujstement.
• Added these parameters to each call on the above-listed functions. It clutters the code a little bit (also the docstrings), maybe we could consider making the cov_model and the known_cov dictionary attributes?
• Added checks at the start of pspec function: asserts if cov_model is a string and if known_cov is given, in case the str isn't among the existing models (['empirical', 'dsets', 'autos', 'foreground_dependent']); if known_cov is given, asserts if the input_data_weight is set to something else than 'identity' - if it is set to identity, then raise warning saying that the input covariance will not be used in the R matrix.

Tested the code with the identity matrix as data covariance.

[adeliegorce]

Running the tests highlighted issues in parsing the covariance information (even for already existing options such as 'dsets' and 'autos'): if input_data_weighting is set to 'iC', for any input cov_model, it is always the empirical covariance that will be computed, therefore the R matrix does not change with cov_model. In particular, it needs to depend on the time_index and so the format of Rkeys needs to be changed.