Change data structure in 'average cosinus' consumer.

[bangerth]

#36 changes the data structure to store previous samples in the spurious autocovariance consumer. We should be doing the same for the average cosinus consumer.

@MantautasRimkus -- feel like giving this a try?

[bangerth]

Also the changes in #38.

[MantautasRimkus]

Got it.