Need to be able to identify probes with low variation

[nhrigby]

Per discussion with @rhiever:
We need a function that identifies probes with low variation among samples, as low variation probes will not be informative downstream.

TBD:

1. What metric do we want to use to examine variation? Beta values?
2. What should the cutoff be for low variation probes? I think we should have some default value, but allow users to specify custom value as well. We can probably use processed samples thus far to help determine appropriate cutoff.
3. Do we want the function to simply return a list/table of low variation probes with their MSE?

[brianchengithub]

1. Beta values is fine.
2. <5% variation could be default. But letting user specify is good.
3. Let's provide a list as part of our set of QC tables/figures.

[marcmaxson]

easily implemented, determine the threshold for cutoff based on a bunch of public data sets.

removing low variation probes is important for prediction, but less important for other science apps.