Implement IO module wrapping JLD2.jl + JSON3.jl

[jc-macdonald]

Summary

Port io.py for saving and loading study results.

Functions to Implement

Function	Description	Julia dep
save_results(path, results::ResultsTable)	Serialize results	JLD2.jl (arrays) + JSON3.jl (metadata)
load_results(path) -> ResultsTable	Deserialize results	JLD2.jl + JSON3.jl

Design Notes

• Python version uses .npz + .json sidecar. Julia equivalent: .jld2 (HDF5-based, handles arbitrary Julia types) + .json metadata.
• Round-trip fidelity is critical: save then load must produce identical ResultsTable.
• Consider also supporting Arrow/Parquet via Arrow.jl for interop with Python/R if the ResultsTable uses Tables.jl interface.

Dependencies

[deps]
JLD2 = "..."
JSON3 = "..."

[compat]
JLD2 = "0.5"
JSON3 = "1"

Acceptance Criteria

• Round-trip test: save_results → load_results produces identical data
• Handles mixed numeric + string metadata
• File format documented