General surrogates

[CommonClimate]

Is your feature request related to a problem? Please describe.

• Various utilities (wavelet, spectral analysis, correlations) use surrogate data (isopersistent, isospectral) in an inconsistent way (e.g. correlations.corr_isopersist() uses correlations.isopersistent_rn() but wavelet or spectral utilities use something else. correlations.py has a red_noise function unused anywhere)
• The Surrogates class currently only has an AR(1) functionality, yet the underlying code for evenly-spaced series (ar1_fit_evenly()) uses an ARIMA model that is much more flexible.
• Phase randomization is only enabled in correlations.phase_ran(), but should be available to other methods.

Describe the solution you'd like

• expose ARIMA parameters at API level
• Enable phase randomization in Surrogates
• Have all MC-based methods use the Surrogates class, not just PSD and wavelets.

Describe alternatives you've considered
Live the rest of my life with an imperfect codebase, moaning in frustration every chance I get.

[CommonClimate]

To do:

• move phaseran() to tsutils
• enable `phaseran() for surrogates
• reform series.correlation() to work like Coherence.signif_test()
• modify correlation unit tests to account for new methods
• same for causality [WON'T FIX]
• enable fBm and power law Surrogates (use tsmodel.parametric_surrogates)
• write unit tests
• implement Surrogates for Coherence objects
• implement Surrogates for Scalograms objects
• Remove stochastic dependency

[CommonClimate]

scalograms and a bunch of unit tests introduced in #534

[CommonClimate]

Updated in #537