CompileML v0.1.1

First complete release. Supersedes 0.1.0, whose wheel carried a stale
hardcoded compileml.__version__ of 0.1.0.dev0 (the PyPI metadata was
correct; the module attribute and artifact compileml_version were not).
The version now has a single source of truth — compileml.__version__ —
which pyproject reads at build time.

Added

• Tuning sweeps (compileml.tune): sweep_whitebox (trees × depth grid with
  measured retention, rank agreement, artifact size, and explanation cost)
  and sweep_bands (band-count grid with retention, Gini gap, and worst
  within-band AUC).
• compileml.bands.band_efficiency: the "money on the table" diagnostic —
  continuous-vs-band Gini gap plus per-band within-band AUC with bootstrap
  CIs and refinable / exhausted / inconclusive verdicts. Validation check 4
  now carries these fields whenever outcomes are supplied, advisory by
  default and gateable via max_within_band_auc=.
• Exact scorecard extraction (compileml.scorecard, stdlib-only): at
  whitebox depth ≤ 2 the artifact collapses into bin → points tables plus
  explicit pairwise interaction grids whose integers re-sum to every
  decision's raw_micro bit-for-bit (score_from_scorecard re-derives any
  decision from the printed tables; asserted in tests). Markdown and CSV
  renderers; compileml scorecard CLI subcommand; refuses above depth 2.
• Docs: tuning guide (howto/tuning.md) answering the configuration
  questions — tree count, depth, the trees-vs-depth asymmetry, band count,
  band efficiency, scorecards — and a FAQ page.
• Decision artifact schema v2: integer-quantized leaves, integer calibration
  tables, fixed-point band ladders, half-micro exact attribution with the
  reconciliation identity, missing-value policy, SHA-256 verify-on-load.
• Pure-standard-library runtime (compileml.runtime): score, band, calibrate,
  explain — enforced stdlib-only by test.
• Compile side: teacher→whitebox distillation, tree extraction for
  scikit-learn / XGBoost / LightGBM (float32 input-precision handling for
  XGBoost), leaf quantization, artifact builder with reason-coverage warnings
  and reproducible (timestamp-free) builds.
• Band builders: quantile_bands, monotone_quantile_bands, and the
  search-and-certify semantic_bands / governance_bands.
• Zero-churn recalibration with predecessor-hash provenance chains.
• Eight-check validation framework running entirely against the artifact
  through the production runtime; CLI exit-code gating.
• Exports: SQL (full pipeline; engine-executed parity tests) and COBOL
  (score + band; artifact integers verbatim).
• compileml CLI: compile, inspect, verify, score, validate, export.
• Benchmarks reproducing every README number; docs site; executable examples.
• Visualization extra (compileml[viz]): payload-driven waterfall,
  decision_drivers, band_drivers, and band_ladder (matplotlib), plus a
  dependency-free waterfall_svg. Plots draw the deployed integers and never
  recompute them; the waterfall's segments are tested against the spec §7.4
  identity exactly.
• Runtime self-check: on artifacts recording exact_attribution, decide()
  refuses to emit an explanation whose residual is nonzero — every explained
  production decision now re-proves attribution integrity in place.
• examples/04_visualization.ipynb: rendered reference gallery for the viz
  suite — arrow geometry, exact remainder truncation, the depth>2 residual
  bar, dependency-free SVG, every colour/sort encoding, per-band facets, and
  restyling. Notebooks can now opt out of CI execution via
  metadata.compileml.ci_execute = false.

Changed

• Documentation doctrine: explain everything by default. Complete attribution
  on every decision is real-time for credit decisioning and makes portfolio
  analytics census-complete; the O(p²) cost is a full-book batch concern,
  which the leaf-time attribution roadmap item targets.