Better swagger management

[ValarDragon]

The current standard for swagger in cosmos seems pretty bad imo:

• We use statik to generate 50k+ lines of code, for a swagger file + 40k+ line openapi.yml we generate
  • Why is this so big? Why does it have so much redundancy?
• This swagger code has to be vendored in every code branch, and included in binaries -- even for nodes with no swagger enabled.
• The resultant swaggerfile isn't even that great. Its a single swagger file thats hard to search, versus a nested file structure.

I'd like to suggest exploring a design space of getting:

• Making the openapi files less massive
• Can we just generate these in nodes that want to serve them, rather than vendor it in the code? Does generating this take too long? Should chains default to having a separate repo for code generating these, and it automatically updates via CI jobs?
• Making more dedicated swagger files per module on top of a unified one

[alexanderbez]

I don't think we haven any one that is super familiar with Swagger at this point. I'm not sure if anyone can even answer all of these questions 😆

[aaronc]

I'm pretty sure we're using an older version of grpc gateway because we're still stuck on gogo proto. Maybe the newer version of gateway does this better?