fft-cnr: FFT-based contrast-to-noise ratio estimation from single frames

Added

• Real-space noise-model detector (estimate_noise_model=True) that fits the photon-transfer relation to flag signal-dependent (shot) noise, returned in a new NoiseModel with a peak_snr method. Significance is calibrated against a pipeline-matched parametric-bootstrap null.
• amplitude_snr derived property on CNREstimate.
• Monte Carlo validation sweep for the noise-model detector (scripts/validate_noise_model.py).
• Weekly CI jobs testing the latest and lowest supported dependency versions.

Changed

• Documented that single-frame quantitative 1/f (spatially correlated) noise correction is unsupported; the spectral NoiseModel fields are reserved sentinels. Correlated noise is mitigated at acquisition (multiple frames, interleaved acquisition, a reference channel).
• Pinned GitHub Actions to commit SHAs and normalized line endings to LF.

Full Changelog: https://github.com/thiebes/fft-cnr/compare/v0.1.1...v0.2.0