Converters: SACC->2pt-FITS + SACC<->OneCovariance

[cailmdaley]

Supply the two glue layers that non-SACC consumers need: a SACC→2pt-FITS converter for the CosmoSIS inference chain, and n(z)→OneCovariance-input plus OneCovariance-output→SACC-covariance glue. Nothing here computes theory or blinds anything; it is theory-agnostic format work.

Desired end state

• A converter turns a {version}.sacc analysis file into a 2pt-FITS file with the exact HDU layout the CosmoSIS chain (including the PSF ρ/τ likelihood modules in the cosmosis-standard-library fork) consumes. The chain runs against the converter's output and produces results identical to running against a directly-assembled file, with no pipeline change. The layout is an 11-HDU multi-extension FITS file; the executable definition of every HDU's columns and header cards is cosmo_inference/scripts/cosmosis_fitting.py.
• Two glue functions bridge OneCovariance: sp_validation n(z) tracers → OneCovariance's per-bin two-column text input (plus its config), and OneCovariance's flat output covariance → a SACC covariance block aligned to the analysis file's insertion order.

Acceptance

• The 2pt-FITS converter is tested against a real reference file and a synthetic fixture (byte-compare where applicable).
• The OneCovariance glue round-trips through a fixture in both directions.

— Fable on behalf of Cail.