fix(seviri): include the first IR row when converting RSR data

[Manny7717]

What

Fixes #253: the SEVIRI RSR converter skips the first IR spreadsheet data row. The issue report was correct: xlrd is 0-indexed, so spreadsheet line 13 corresponds to start_rowx=12, not 13.

For IR7.3 in the bundled workbook, row 12 starts at wavelength 6.35, while the current loader starts at row 13 and incorrectly begins at 6.37.

Change

In rsr_convert_scripts/seviri_rsr.py, the IR-channel branch now reads rows 12..111 instead of 13..112 for the wavelength column and every Meteosat response column.

Verification

• Workbook proof from pyspectral/data/MSG_SEVIRI_Spectral_Response_Characterisation.XLS:
  • row 12 col 0 = 6.35
  • row 13 col 0 = 6.37
• New regression test test_ir_channels_include_the_first_spreadsheet_row
  • fails before the fix: loaded first wavelength is 6.37
  • passes after the fix
• Test runs:
  • .venv/bin/python -m pytest pyspectral/tests/test_seviri_rsr.py pyspectral/tests/test_config.py -q -> 4 passed
  • .venv/bin/flake8 rsr_convert_scripts/seviri_rsr.py pyspectral/tests/test_seviri_rsr.py pyspectral/tests/test_config.py -> passed

Why this is scoped correctly

This change only touches the IR-specific row window in the shared loader path. The non-IR channels already use start_rowx=12, and HRV uses a different block entirely.

[Manny7717]

pre-commit.ci was right — my new test file was missing module/function docstrings (D100/D103 under flake8-docstrings). Added them in 7eb0722 and verified locally with the exact hook environment (flake8 + docstrings + bugbear + debugger): clean, and the regression test still passes.

[djhoese]

@adybbroe will need to be the final say on this pull request, especially since it will mean regenerating the RSR data and uploading it to the LUT data host. I think this looks good from what I understand except that this would be the first RSR script with tests. These aren't really user facing and are run "once" so we don't typically have tests for them. I almost want to say the test module should be removed, but @adybbroe should make the final call on this too. Thanks for fixing this @Manny7717.