deprecated transiting_code, legacy_pspec.py and C_empirical()

[nkern]

This deprecates the transiting_code folder that has been hanging around during the graduation of OQE. This is now completed and therefore is no longer necessary to keep around.

This also deprecates the legacy_pspec.py module that was put in place in case other users wanted to interface with legacy techniques through hera_pspec. This was never really used (according to my knowledge), so we are deprecating it to clean things up.

Lastly, we found that within pspecdata, the C_empirical() and C() methods were doing the same thing of calling the utils.cov() function to calculate empirical covariances. We are deprecating C() to avoid confusion and ambiguity as to what function is called to calculate an empirical covariance.

Addresses #93

[coveralls]

Coverage increased (+4.4%) to 96.713% when pulling 33e234e on deprecate_legacy into 19db6ce on master.

[philbull]

Looks good. Only very minor comments about this actual PR, but it raises some interesting conceptual issues that we should address soon.

[philbull]

Maybe model= instead of mtype=?
I'm a bit worried about consistency here by the way. What happens if the user calls this with one model type, and then tries to use another model type later on? Is the cache dropped and recomputed in this case? Or should we store one cache per model type?

[philbull]

It's also unclear to me how we will support more sophisticated covariance matrix methods in future. Should we be getting the user to use set_C manually? Or should we attach some wrapper object, similar to PSpecBeam, that handles all of the covmat stuff?

[nkern]

Yeah that's a good point about the cache. As it stands there is only one slot for a covariance matrix, regardless of its type, so the most recently "set" covmat is the one it will pick up on. Perhaps when we re-haul inverse covariance weighting we should take a closer look at how we implement the storing of various covariance matrix types.

[philbull]

Looks good to me!