Added Digital Compression

[clackner-gpa]

This adds a new compression option for datablobs.

It is used if < 10% of points represent actual changes in values (e.g. Digitals)

It compresses the data into a byte array containg:
# of sampes (int)
SeriesID (int)
First Timestamp (long)
Ticks between samples (long)
Offset and scaling for values (2 doubles)
And the data for every change in Value in form
Timestamp (short long) Value (ushort)

With this PR there is no difference except in length of the data blob.

[StephenCWills]

I wonder if rounding errors could potentially introduce additional data points between state changes. Do you think it might cause problems downstream if points.Count > m_samples?

[clackner-gpa]

m_samples get's removed from the blob so I am not sure that even matters.

By definition we assume these "Digitals" are evenly sampled so it will generate an evenly sampled flat line until it hits a state change. In theory it is possible that we Loose some points in between or generate extra ones, but since it's a flat line anyway I am not too concerned about that.

Occasionally we may compress a latched analog this way, which I suppose could be an issue, but I believe we already assume fixed sampling rate in a few places when doing math so not sure that really matters either.

[StephenCWills]

Off the top of my head, my main concern would be this.

openXDA/Source/Libraries/FaultData/DataAnalysis/DataGroup.cs

Lines 323 to 331 in f0ccd33

	// If the data being added matches the parameters for this data group, add the data to the data group
	// Note that it does not have to match Asset
	if (startTime == m_startTime && endTime == m_endTime && samples == m_samples)
	{
	m_dataSeries.Add(dataSeries);
	return true;
	}
	return false;

If the actual number of decoded data points doesn't match DataGroup.m_samples, then the DataGroup.Add() method will fail to even include the digital series in the DataGroup. This is called directly by DataGroup.FromData() so the decoded data would just be silently dropped on the floor, and we would be wondering why it's not getting returned to the visualization.

Also, I suppose I'm assuming that there will be a follow-up PR that includes changes to DataGroup.FromData()?

[clackner-gpa]

Note with the switch to use ulong for Timestamps, we add an extra byte per state change. We could further reduce this and optimize for space, but I don't think it's worth it considering there will be very few state changes in a true digital. Anything we save over a full standard blob is already a bonus.