feat: subtract sample fluorescence on absolute 1-D SAXS profiles

[D-sudoasd]

Problem

Metal SAXS at energies above absorption edges carries sample-generated X-ray fluorescence. Empty-cell/NIST-blank and BioSAXS buffer subtraction do not remove it. Users needed an opt-in, fail-closed 1-D correction that records provenance.

Cause

The package had no fluorescence ledger token or kernel. Detector-space blank math is the wrong place for an isotropic 1-D additive term.

Fix

• New kernel subtract_fluorescence on absolute cm^-1 data: I_corr = I_abs - beta F(q)
• Methods: constant, high_q_mean, high_q_median, measured_profile
• Ledger token fluorescence; unknown u(F0)/u(beta) stay NaN; negatives are reported, not clipped
• CLI saxsabs subtract-fluorescence
• Workbench Tab 3 after optional buffer; Tab 2 1-D I(q) (not I-chi)
• Optional BL19B2 integrate1d post-step
• Default off; NIST 2-D blank formula unchanged; paper.md unchanged
• Sequential buffer then fluorescence keeps buffer statistical extras and NaNs unknown combined terms

Verification

• pytest tests/test_fluorescence_subtraction.py tests/test_intensity_state.py tests/test_cli.py tests/test_buffer_subtraction.py tests/test_workbench_scientific.py tests/test_bl19b2_integrate1d.py tests/test_detector_reduction.py -q: 238 passed
• Full pytest -q: 785 passed
• ruff check src tests: clean
• Real CLI (this repo src on PYTHONPATH): two identical subtract-fluorescence --method constant --f0 2.0 runs with JSON method/f0/points