[entropy_src] Statistical Arguments on entropy_src health check definitions

@cfrantz is already working on this.

Since Jan 12, 2024.

I'm approving #1943, pending a more detailed review of the statistical arguments around how to define health checks when merging several (assumedly) independent data streams. Any one with strong opinions on statistics should weigh in on this mathematically motivated point. This discussion should at least include: @zi-v, @martin-lueker, @felixmiller, @cyrilguyot. Everyone should feel free to weigh in.

This deliverables should be at least.
- a detailed definition of the test definition
- A formula relating the RCT and APT thresholds, queue depths, and the nominal miniumu entropy/bit to the rate of false positives in the entropy/bit of an given source.
- Guidelines for setting the thresholds
- An outline of the assumptions: Notably: "How weak a correlation can we tolerate between analog channels?"
- Limits of the entropy/bit for the interleaved entropy stream, given our assumptions about the individual entropy streams.
- A second review of the RTL, confirming that the implementation matches the more clearly defined test.