add humanize function to generate pink noise

[michiboo]

Closes #56

[Datseris]

Hi, before reviewing the details, this needs some work at a higher level.

1. a function humanize! should add the noise components to the positions (or even velocities) of existing notes.
2. You have to check that the standard deviation of the noise process is sufficiently small, and can be controlled by the user. What is best is that the user provides a σ, and the resulting noise is normalized to have standard deviation σ.
3. The user should be able to propagate d, φ parameters to ARFIMA.jl.
4. Please write unit tests.
5. Have you confirm that the timeseries you get from ARFIMA have power-law spectra?

[michiboo]

@Datseris Hi I have coded a test, no.3 and function to add the noise components to the positions. But I am struggling to get a power-law spectra.

[Datseris]

Hi, yeap I was fearing that we won't be getting power-law spectra (at least not crisp clean ones) with this approach. Here is what I propose:

0. Notice that humanize modifies the given notes. Thus, it should end in ! in accordance with the Julia style guide. Rename humanize to humanize!.
1. Change the call signature of humanize to: humanize!(notes, property, σ, noisetype, args...). The property can be either :position, which does what you do now, or :velocity, which adds noise to the note velocity. Instead of doing notes[i].position = ... you can do setproperty!(notes[i], property, ...) instead, and you cover both lines with 1 line of code!
2. The noisetype is a symbol that configures what kind of noise we add. This allows us in the future to add different type of noise, that perhaps is more like a power-law.
3. The args... are propagated into the function that gives you the noise, and of course have to be compatible with the chosen method. In our case, where method = :ARFIMA, we would have noise = arfima(length(notes), σ, args...).
4. Please ensure that the σ the user gives is the σ the method gets, i.e. transform the output noise to noise = σ .* noise ./ std(noise).
5. Add a test about the velocity setting as well.

Then this is good to go.

P.s.: If you ever find some code that generates power-law correlated noise please let me know! I still haven't found anything!