Add optika.systems.LinearSystem, a linear forward-model approximation to SequentialSystem

[roytsmart]

Summary

Adds optika.systems.LinearSystem (and its abstract base AbstractLinearSystem), a fast linear forward model that approximates a SequentialSystem.

Instead of raytracing every scene, a LinearSystem is characterized once by a set of precomputed models:

• a distortion model mapping object-plane coordinates onto the detector,
• an area_effective (effective area) model,
• optional vignetting and field_stop models,
• and a sensor.

image() maps a scene's energy spectral radiance to detector electrons by conservatively regridding it through the distortion model, weighting each cell by the effective area, vignetting, and field stop, then applying the sensor's exposure model. The expensive conservative-regridding operator is built once in weights() and reused by image_from_weights(), so many scenes can be imaged through the same optics cheaply.

Supporting changes

• Shaped on the calibration models. AbstractDistortionModel, AbstractVignettingModel, and AbstractEffectiveAreaModel now inherit the Shaped mixin, with shape implemented on the concrete models (SimpleDistortionModel, PolynomialDistortionModel, PolynomialVignettingModel, InterpolatedEffectiveAreaModel). Each returns the model's batch shape (excluding the wavelength/field domain axes), which is what lets LinearSystem.shape report the system's parametrization.
• Moved Plottable/DxfWritable from AbstractSystem to AbstractSequentialSystem. A LinearSystem has no geometry to plot or export to CAD, so the common AbstractSystem base no longer forces those mixins; the geometric SequentialSystem keeps them.

Tests

_linear_test.py adds TestLinearSystem with two fixtures (a minimal system, and one with vignetting and a field stop), driving image() end-to-end with noise on and off, plus the model interface. The distortion/vignetting/effective-area test bases gained shape coverage.

Docs

LinearSystem has a runnable .. jupyter-execute:: example that assembles the component models and images a USAF-1951 target, mirroring the SequentialSystem example. AbstractLinearSystem documents the full model, and a Features bullet was added to the landing page.

Follow-ups (known limitations)

• A scene with only a single wavelength cell currently fails in image(): the size-1 wavelength axis is dropped and the regridding cannot align the weights.
• LinearSystem.image expects energy spectral radiance, whereas SequentialSystem.image takes photon radiance; the two image() methods disagree on units.

🤖 Generated with Claude Code

https://claude.ai/code/session_01CtdKmedevWkDab6BWupXqQ

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.55%. Comparing base (bf75a1f) to head (6381076).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #152      +/-   ##
==========================================
+ Coverage   99.54%   99.55%   +0.01%     
==========================================
  Files         116      118       +2     
  Lines        6374     6553     +179     
==========================================
+ Hits         6345     6524     +179     
  Misses         29       29              

Flag	Coverage Δ
unittests	99.55% <100.00%> (+0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.