[Feature] Simplify the multi module Agentic Subprocess Repo

### Description of Problem

Repo : https://github.com/finos/fluxnova-plugins/tree/main/agentic-subprocess
PR comment : https://github.com/finos/fluxnova-plugins/pull/29

Is it possible to combine the modules into two instead of so many which is easier to understand, release and modify. For a real Fluxnova/AI integration we probably need all of them

one module for the entire plugin like how it is laid out for MCP plugin, which is responsible for parsing the BPMN, discoveing the tools, invoking them and orchestratid
Second module which just integrates with LLM - This is what I hope we modify for integrating with our LLM

### Potential Solutions

Refactor the Agentic plugin code into two modules for better long term maintainance
one module for the entire plugin like how it is laid out for MCP plugin, which is responsible for parsing the BPMN, discoveing the tools, invoking them and orchestratid
Second module which just integrates with LLM - This is what I hope we modify for integrating with our LLM

Start by reviewing the agentic-subprocess repository tree and the discussion in PR 29. Inventory the existing modules and how parsing, tool discovery, invocation, orchestration, and LLM integration are divided. Done means the plugin is reorganized into one module for core orchestration and one module for LLM integration, with the resulting structure easier to understand, release, and modify.

Written by the indexing model from the issue text.