feat: TritonCore DistributedRuntime Worker Example

[rmccorm4]

Example:

• Following from hello_world example - client/server send strings back and forth as text in/text out, and passes through to a mock LLM python model served by TritonCore
• Uses TritonCore Python API directly rather than TritonCoreOperator - see details below

Things to iron out:

• Contract on request object type/schema between "client" and "server" for the RequestHandler / @triton_worker
  • Seems it can be any arbitrary object as long as they agree with each other / can handle it
• Flexibility of TritonCore object
  • Similar to request object schema - need some information/definition on request object to flexibly handle arbitrary models served by TritonCore
• Re-use of TritonCoreOperator objects
  • Has certain expectations around request plane usage and remote/local request/response that may need to be refactored/separated out when being used behind Rust DistributedRuntime which may have already done the request plane / data movement logic (remote vs local)

[nnshah1]

nice!