band_distribution normalization tables

Precomputed, interpolated normalization tables for the band_distribution library: the integral of the nuclear-recoil (NR, PpqN) and electron-recoil (ER, PpqG) band probability densities over a fit window in (Ep, Eq), tabulated over the physics parameters an MCMC varies. Computing this integral directly takes ~1–2 s per call; interpolating a table takes tens of microseconds. Each fit window has its own pair of tables; file names carry the band and window, norm_<band>_ep<Ep range>_eq<Eq range>.h5 (ranges in keV). New windows are added as new versions of this record, so the concept DOI always resolves to the full, latest set. Windows Ep 2.5–350 keV, Eq 0.75–200 keV (added in version 1.0) norm_NR_ep2.5-350_eq0.75-200.h5: NR table over (k, F0, V, p0, dp = p10 − p0, q0, dq = q10 − q0), 8×3×4×4×3×16×4 = 73,728 Chebyshev–Lobatto nodes; Z = 32, eps = 3.0e-3 fixed. norm_ER_ep2.5-350_eq0.75-200.h5: ER table over (F0, V, p0, dp, q0, dq), 3×8×4×5×10×9 = 43,200 nodes; eps = 3.0e-3 fixed. plan_*.json: the node count measured for each axis (Chebyshev coefficient tails, target 1e-7 per axis) and the parameter box. build_log_*.txt: build output, including validation against 200 held-out, directly computed points per table. Parameter box (both tables): k 0.13–0.22 (NR only); F0 1e-5–1 (NR), 0.1–0.35 (ER); V 2.7–3.3; p0 0.01284–0.1156; dp 0.18–0.59; q0 0.04744–0.4; dq 0–0.353. The tables never extrapolate: a query outside the box raises an error. Accuracy (relative interpolation error at 200 held-out points, acceptance threshold 1e-6): NR worst 1.22e-7, rms 4.0e-8; ER worst 2.74e-7, rms 6.9e-8. Each integral was computed to relative tolerance 1e-7. Provenance: built with band_distribution commit 457ff90 (slurm/build_table.sh), integrals evaluated in the band_distribution container image built at commit a531e74; the region, box, fixed parameters, node counts and commit are also stored as attributes in each HDF5 file. Use PpqPDF(ep_min, ep_max, eq_min, eq_max, ep_data, eq_data, ppqn_table="norm_NR_...h5", ppqg_table="norm_ER_...h5") in band_distribution's python/ppq_pdf.py, or fetch by name with python/normtables.py.