The CapNInterdigitalTee cap-matrix dataset was an exact mirror of the NCap cap-matrix dataset prior to PR #15 (only the design.coupler_type string differed). This PR applies the same 118-record Sobol + lattice augmentation that was merged into coupler-NCap-cap_matrix.json in #15, keeping the two datasets in sync.

Contents:
- existing 430 records: preserved byte-for-byte from pre-PR revision 6b4001554a05
- appended 118 new records: identical geometry and sim_results to the new rows in PR #15, with design.coupler_type rewritten to 'CapNInterdigitalTee'

Validation (same script used in PR #15 review):
- Schema OK: all top-level keys, sim_options (byte-identical), sim_results keys, units=fF, coupler_type=CapNInterdigitalTee all match existing records
- 118/118 new records have finite, positive capacitances
- Nearest-K (K=5) median percent diff vs existing neighbours: 1.4% - 4.8% per cap
- Two-sample KS tests: all 6 capacitance distributions within-distribution vs existing
- Only cosmetic finding: design.design_tool is 'Qiskit Metal' for the new rows (canonical casing used by modern SQuADDS simulations/result_normalization.py) vs 'qiskit-metal' in legacy rows; not used as a filter key anywhere in the SQuADDS API