Need to add APD to algorithm extension

[MichaelCotton]

with options for downsampling

[aromanielloNTIA]

Working on this now. Here are some thoughts and questions:

• I think it makes sense to add a ProbabilityDistributionAnnotation segment, rather than a specific APD segment. This is general enough to support other types of distributions which may be generated. It might look like:

name	required	type	unit	description
function	true	string	N/A	The estimated probability distribution function, e.g., "cumulative_distribution", "probability_density", "amplitude_probability_distribution"
units	true	string	N/A	Data units, e.g., "dBm", "volts", "watts".
probability_units	true	string	N/A	The unit of the probability values, generally either "dimensionless" or "percent".
number_of_samples	false	integer	N/A	Number of samples used to estimate the probability distribution function. In the case of a downsampled result, this number may be larger than the length of the data.
reference	false	string	N/A	Data reference point, e.g., "signal analyzer input", "preselector input", "antenna terminal".
probability_start	false	double	probability_units	Probability of the first data point.
probability_stop	false	double	probability_units	Probability of the last data point.
probability_step	false	double	probability_units	Step size between data points. This should only be used if the probability step size is constant between all data points.
probabilities	false	double[]	probability_units	An array of the probabilities for all data points. This must be used if the probability step size is not constant (i.e. if the probability axis cannot be specified fully using probability_start, probability_stop, and probability_step.

• I will need to think some more on the best way to document downsampling, since there are multiple ways to do it. It's almost like the algorithm annotation needs an algorithm annotation... but let's try to avoid that rabbit hole.
• Other than the APD/CDF/PDF, which distributions may be common for RF analysis? Keeping them in mind would be helpful to make this as general as possible.

[aromanielloNTIA]

Adding to the above definition for a ProbabilityDistributionAnnotation segment, to account for downsampling. After thinking through a few ways to do this, they all result in adding tons of new keys to describe a variety of possible ways to downsample the data. I think that a single string containing a human-readable description of the downsampling applied would be a better solution.

name	required	type	unit	description
downsampled	false	boolean	N/A	Whether or not the probability distribution data has been downsampled
downsampling_method	false	string	N/A	The method used for downsampling, e.g "Uniform downsampling by a factor of 2" or "Distribution binned by amplitude with a resolution of 0.5 dB"

[aromanielloNTIA]

Implementation differs from what is described above, but the Graph object added in #42 supports metadata for (optionally downsampled) APD results. This issue should be closed as resolved when it is merged.