healthchecker - Enhance current performance analysis for disk write queue, add factors of item count, disk performance and time span

Perry suggests:

-If it's possible, teaching the user to "fish" by showing them exactly where to spot this data themselves would be helpful as well. One idea would be to generate an actual representation of what the UI would look like were they able to see this. Another would be generate a hyperlink that actually took them to the correct graph and timescale for a various stat or analysis or alert. I personally (among others) spent a lot of time getting the UI into a place where it was really usable to identify at least the high level problem points without having to go down to the command-line and look at 10 nodes at once. It would be great if we could teach our users how to use it in the same way. I can honestly tell an immense amount of information about a cluster's health just be looking at the "summary" tab. That comes with experience certainly, but hopefully we can automate and explain it as well. For example, correlating things like a customer saying "it seems to be running slowly" into "you have a lot of disk reads" into "you have a high cache miss ratio" into "you have a low resident item ratio" into "you have a high mem_used" into "your items have been growing by 10M every day for the last 5 days"..."AND THAT'S WHY IT'S SLOW" is going to be a great leap forward. Another good one is "it seems to be running slowly" into "you have a lot of disk reads" into "you have a high cache miss ratio" into "you have a low resident item ratio" into "you have a high mem_used" into "your item count seems very high relative to where it was yesterday" to "your sets/sec just jumped 30x in one minute"...you get the picture. It's easy-ish to point out these kinds of things after you've already seen them, it's much harder to predict what's going to happen before we see it.