g-adopt/srts: v0.1.0 — Initial Python release

First release of srts as a Python package.

This replaces the original Fortran/C pipeline (dofilt_ES_new) with a pure Python implementation built on NumPy, SciPy, pyshtools, and h5py. The filtering methodology and resolution operators are identical to the original; all outputs agree with the Fortran pipeline to 6–7 significant digits.

The package supports S12RTS, S20RTS, and S40RTS resolution operators and provides two interfaces: a file-based pipeline function for workflows where data lives on disk as depth-slice ASCII files, and a class-based API for in-memory use (e.g. directly from a G-ADOPT simulation).

The SH expansion in expand_batch uses a fully batched GEMM path, giving roughly 10–16x speedup over the equivalent per-layer loop depending on grid size.

Model data (~4.8 GB for all three operators) is downloaded automatically on first use from a public storage bucket.