LLM Chain: Add sequential multi-step LLM call execution

[vprashrex]

Is your feature request related to a problem? Please describe.
Kaapi has no native way to build multi-step AI pipelines like Voice-to-Voice with file search, AI Assessments, or Translation Pipelines, each step has to be a separate manual API call.

Describe the solution you'd like
Create llm_chain as workflow manager that orchestrates sequential llm_call executions, each block's output automatically becomes the next block's input, with checkpointing, intermediate callbacks etc.

Describe alternatives you've considered
A clear and concise description of any alternative solutions or features you've considered.

Additional context
llm_chain should adds zero LLM logic, it only manages at the orchestration level, so any provider, config, or input type that works in /llm/call today or in the future works in /llm/chain with no changes.