MAINT deprecate `self.average` 1D

in preparation for the JTFS and 0.4 we could deprecate average in Scattering1D.

While reusing the 1D frontend for frequency scattering it came to our attention that we could utilize T to define the semantics of frequential averaging. This would evade introduction of an additional attribute F. Moreover, T could potentially cover the semantics of averaging, allowing for removal of average:

• T=None will perform default mode averaging, i.e. T=2^J
• T=0 will correspond to average=False
• Any other integer T will stride over the specified support

[janden]

We should also think about what this implies for 2D. Right now, we don’t have an equivalent for T there but this may change.

[lostanlen]

I think it’s necessary to have an F parameter in the JTFS frontend, but this can be passed to the Scattering1D frontend for frequential averaging as T?

F will be passed as T to the frequential scattering_filter_factory, yes. But there won’t be a Scattering1D object for frequential scattering, because @eickenberg doesn’t want objects inside other objects. Rather, the JTFS object will inherit and extend from time Scattering1D.