Dr. Elephant Configurable Severities

Nobody has claimed this yet.

In our last weekly meeting, we had a discussion on renaming the severity levels in Dr. Elephant. For details take a look at the agenda [here](https://docs.google.com/document/d/17PGHbWXY_-W9AW4Nb-2qGz8twBJOfJvUqD-iyAD900U/edit?usp=sharing) and the meeting minutes [here](https://docs.google.com/document/d/13-kPn__2By8X4FpnroAus_s_ZxThO8Ew6elFCA8bBiE/edit?usp=sharing).

We discussed some concerns over renaming them like
1. Services depending on Dr. Elephant severities via the rest end point might break.
2. It can be confusing for regular users of Dr. Elephant if we change the severity naming convention.

We finally concluded on making the severities configurable so that people can decide for themselves on what type of convention they want to follow. Of course, we will maintain a standard default value.

As part of this goal, we would love to see the severities to be configurable in terms of
1. The severity naming convention. This will give flexibility for each org to have their own severity naming convention. The severities could have names or even star ratings.
2. The number of severity levels. Currently we have 5 threshold levels(none, low, moderate, severe and critical). This could be made configurable to have 2 or 3 or even 10 levels.

In addition, it would really good if we can map severity levels to heuristics. Example, for some heuristics like "Exception" or Configuration check type of heuristics the outcome could be simply good or bad (true/false) and in such cases it makes sense to have only 2 severity levels and in other cases we could use the regular 5 or 10 level thresholds.

We welcome interested people from the community to discuss and take up this issue.

Start by reading the linked agenda and meeting minutes, then trace where the current five severity thresholds, names, and heuristic outcomes are defined. Define the configurable levels, naming conventions, heuristic mappings, and default behavior while checking that REST endpoint consumers remain compatible; the issue names no files or tests to run.

Written by the indexing model from the issue text.