Improve and refactor Build Flow

[Urigo]

Improving the process of creating artifacts and deploying them.

Today we have the option to separate the building of the handlers' artifacts and loading them to the gateway.
The artifacts that can be generated in two different formats - JSON or Code Files

So when developers use a CDN (like GraphQL Hive) to load and store these artifacts, the process is:

1. Build artifacts
2. Upload
3. Use it in the configuration

With these configurations, Mesh doesn’t build artifacts again on the gateway.

All this is great but currently these types of configurations are different for each handler:

• For GraphQL handler, they should give introspection

• For JSON Schema/OpenAPI/RAML, you provide a bundle and it doesn’t matter which handler you use since they are all the same since bundle already has all the configuration

• No option for gRPC

• No option for Postgres, etc..

• Consider a common option like artifactsPath instead of the specific options above

Also now that we have that separation, we could also:

• Split handler implementations to reduce the runtime code into (1. Validation 2. Artifacts generation 3. Execution)

[ardatan]

Now we have guides for Hive and different SDL-based approaches in OpenAPI handler etc.