SPAR reviewer-priority validation and provenance v2.1.0

Data and reproducibility package for the revised manuscript “Support-Preserving Sparse-to-Dense InSAR Forecasting with Anchored Neural Residuals”. Version v2.1.0 adds the reviewer-priority validation suite for SPAR: five-partition direct raw-observation evaluation, frozen multi-seed confirmation, Nadeau–Bengio corrected resampling statistics, training-count-matched 100 m and 2 km buffer controls, external-tile evidence, and known-truth plus deliberately leaky interpolation-target controls. The package includes the EGMS source data used by the manuscript, a compact synthetic test case, legacy result summaries retained for traceability, the complete GitHub release archive, and a machine-readable release manifest. The frozen release is GitHub commit ea860d527ae19fbde51cdb791547d82ca8556593. Archive SHA256: 7e7fc8b01a8a7fb6766f51ff807abdb49ff80a74274360bd7f26a0ed52c0485e. Current source hashes support reproduction of the released workflow; exact launch-time extension-source hashes for the original frozen SPAR runs were not captured and are therefore disclosed as unavailable historical provenance. Transformer, STGCN, graph-model, and RSASE experiments are outside this release. Redistributed EGMS content should be attributed to European Union, Copernicus Land Monitoring Service information; https://egms.land.copernicus.eu/.