Phase 1A: Core infrastructure — registry, model, controller skeleton

[lg345]

Parent: #82

Scope

Build the foundational infrastructure that all subsequent phases depend on.

Deliverables

1. Module layout — Create the directory structure: model/, analysis/, controller/ with __init__.py files.

2. model/run.py — Refactored spectroscopy_run with:

   • self.results = {} (flat dict, dot-separated keys)
   • self.update_status() preserved
   • Data loading methods (load_run_keys, load_run_key_delayed) retained but writing to results dict

3. model/experiment.py — experiment and spectroscopy_experiment classes moved as-is.

4. analysis/registry.py — The step registry:

   • @register_step(name) decorator
   • @register_reduction(name) decorator
   • get_step(name) lookup function
   • get_reduction(name) lookup function
   • list_steps() for introspection

5. controller/config_parser.py — YAML parser that:

   • Reads a YAML file
   • Validates required sections exist (experiment, data, pipeline, output)
   • Returns a structured config object

6. controller/pipeline_runner.py — Step dispatch loop:

   • Iterates pipeline section
   • Looks up each step name in registry
   • Calls step with (run, **args)
   • Logs execution

7. controller/batch_manager.py — Shot chunking and multiprocessing:

   • Breaks runs into shot ranges (existing break_into_shot_ranges logic)
   • Manages multiprocessing.Pool
   • Reconverges batch results into run-level list

8. Pipeline class — The user-facing entry point:

   • Pipeline.from_yaml(path) class method
   • .run(cores, batch_size) method
   • .results attribute for post-run access

Acceptance criteria

• Pipeline.from_yaml("test.yaml").run(cores=4, batch_size=1000) executes without error on a minimal YAML with placeholder steps
• Registry correctly discovers and dispatches decorated methods
• Batch manager splits and reconverges shot ranges correctly