Experimental representation analyzer

## About

We're creating an experimental analyzer type that uses stored representations. The aim on Wed is to just get something live end-to-end and see if the whole thing makes sense.

## @ccare

- We need to deploy the `add-representers` branch of the orchestrator. This calls out to the existing analyzers in the normal way, but also calls out to a script called `generate_representation`, that takes the same inputs as `analyse_iteration` and also writes to a file in exactly the same way as the the analyzers. The file will be called `representation.txt`, and is the equivalent of `analysis.json` if that is important for you to know.
- The `application_name` in secrets.yml MUST be changed before this is deployed, else we'll start cannibalising the real requests to the real analyzers.
- We need an RDS set up that the analyzer can talk to and some params setting in the secrets.yml (see [here](https://github.com/exercism/analyzer-orchestrator/pull/29/files#diff-53ca74e31dff0b0fde63a83bbd9f4cfdR11-R14) for what needs adding).
- We will then need to deploy the things that DJ and Erik make. These will be pretty much identical to the analyzers from an external interface POV, and have exactly the same risk-profile as them.

## @sleeplessByte @ErikSchierboom

You are both going to create a `representer`. You have a [JS repo](https://github.com/exercism/javascript-representer) and a [C# repo](https://github.com/exercism/csharp-representer) to use.

The Representer:
- Has the same inputs as the analyzer.
- Creates a representation.txt (rather than an analysis.json)
- The representation.txt should contain an AST representation of an iteration, in whatever form is the most compact (we're looking up by mysql text comparsion, so compressing is good).

Once we've got a first cut of the bits, we can start making it more clever/complex :)