advanced-topics/troubleshooting: add section "Diagnosing problems with metrics"

[odeke-em]

Following suit with #472

Borrowing an idea from the MySQL website "Performance schema quickstart" https://dev.mysql.com/doc/refman/8.0/en/performance-schema-quick-start.html
I just saw the value of having a direct advisory of how to use a solution to diagnose and solve problems.

It would be useful for us to make a section titled "Diagnosing problems with metrics" and link to that in the same fashion at the top of the "Metrics" introduction page.
This ensures that the practicalities of using metrics will be properly communicated and prescriptive, thereby creating a direct use for the content on the website.

Perhaps this is where our collaboration with an SRE would be useful and we could take a page out of @oaktowner's presentation at Next 2018 where he collaborated with a veteran Google SRE Nori Heikkinen as per https://www.youtube.com/watch?v=XPtoEjqJexs

Their presentation style articulated why you need Istio, how and what it provides to fill in the golden signals emphasized in the SRE Handbook https://landing.google.com/sre/sre-book/toc/index.html

Such as section would be very informative and directly add value to the project by showing a direct prescription instead of letting people have to think of when to apply or repeatedly explain why they'll need OpenCensus.

/cc @rakyll @bogdandrutu @shahprit @hvent90

[hvent90]

If someone who has the real-world experience of using metrics in a site-reliability capacity is able to afford the time to detail a high-level real-world implementation, then I can go in and write the guide.

Context: I am familiar with the API accross java/go/node/python, however I haven't used it in a real-world scenario -- so my implementation isn't battle tested and would be flawed.

Another scenario is I could just write the section to get it out the door and then a more knowledgable individual polishes it up later.