NeuNorm 2.0.0

NeuNorm 2.0 is a complete, scipp-based rewrite of the
library. It is a breaking change: code written against the 1.x
NeuNorm.normalization.Normalization API will not run unchanged. See the
1.x → 2.0 migration guide, and pin NeuNorm<2 to stay on the
legacy API.

Added

• Scipp-native processing. All data are scipp.DataArray objects that carry
  variances, so uncertainty is propagated automatically through every step.
• Time-of-flight (TOF) support for the VENUS pulsed source — wavelength-resolved
  (hyperspectral) transmission T(λ, x, y), TOF binning, and histogram rebinning.
• Event-mode processing for Timepix3 detectors: NeXus/HDF5 event loading and
  pulse reconstruction (neunorm.loaders.event_loader, neunorm.tof).
• End-to-end detector pipelines in neunorm.pipelines, one per
  detector/facility combination: run_mars_ccd_pipeline, run_mars_tpx3_pipeline,
  run_venus_ccd_pipeline, run_venus_tpx1_pipeline,
  run_venus_tpx3_histogram_pipeline, and run_venus_tpx3_event_pipeline.
• Composable processing functions for building custom workflows: dark
  subtraction, reference (open-beam/dark) preparation, transmission normalization
  with proton-charge correction, ROI clipping, run combination, air-region
  correction, spatial rebinning, and Poisson/systematic uncertainty helpers
  (neunorm.processing).
• HDF5 as the primary output format (neunorm.exporters.hdf5_writer.write_hdf5),
  with detector masks and provenance metadata; TIFF export retained as secondary
  (neunorm.exporters.tiff_writer.write_tiff_stack, via scitiff).
• Loaders for TIFF, FITS, NeXus event, and NeXus metadata
  (neunorm.loaders), including shutter-count and TOF-spectra readers.
• Resonance / Bragg-edge analysis and TOF statistics (neunorm.tof.resonance,
  neunorm.tof.statistics_analyzer).
• Pydantic v2 configuration models (e.g. BinningConfig, EventData) for
  explicit, validated configuration.
• Sphinx + autodoc documentation published at
  neunorm.readthedocs.io, with per-workflow guides
  and a full API reference.

Changed

• Import name is now neunorm (lowercase), not NeuNorm. The PyPI/conda
  distribution name remains NeuNorm, so pip install NeuNorm is unchanged, but
  import NeuNorm becomes import neunorm.
• Minimum Python is now 3.11.
• Development uses pixi (pyproject.toml [tool.pixi.*] +
  pixi.lock); the 1.x conda environment.yml / conda.recipe are retired and
  archived under archive/neunorm-1.x/.
• Optional features are exposed as extras: viz (plopp/matplotlib) and
  performance (Numba acceleration); Numba is optional and degrades to a no-op
  when absent.

Removed

• The entire 1.x stateful API, including NeuNorm.normalization.Normalization
  and NeuNorm.roi.ROI. There is no drop-in compatibility shim — the flat-field
  normalization physics is preserved, but the API is new. The 1.x source is kept
  for reference under archive/neunorm-1.x/.

Migration

See the 1.x → 2.0 migration guide for a step-by-step mapping
from the legacy Normalization workflow to the 2.0 pipelines and composable
functions.

Install: pip install NeuNorm==2.0.0 or conda install neutronimaging::neunorm