extract from imagehdu so that full fov is covered

[oczoske]

This fixes #211, compare
micado_vs_psf_ncpa_ycut_correct.pdf to the figure linked there. Instead of looking at the pixel edges I replaced round() by the more appropriate floor() and ceil() functions and made sure that the proper pixel range is included.

[oczoske]

Incidentally, extract_area_from_imagehdu relies on functionality that requires overhaul in connection with the off-by-one error. imp_utils.header_from_list_of_xy seems dodgy as well, though apparently not critical.

[oczoske]

Oh dear, failing tests. These need to be sorted.

[codecov]

Codecov Report

Patch coverage: 100.00% and project coverage change: -0.26 ⚠️

Comparison is base (79d0aea) 75.41% compared to head (4566eea) 75.15%.

Additional details and impacted files

@@              Coverage Diff               @@
##           dev_master     #238      +/-   ##
==============================================
- Coverage       75.41%   75.15%   -0.26%     
==============================================
  Files             149      152       +3     
  Lines           15283    15668     +385     
==============================================
+ Hits            11525    11775     +250     
- Misses           3758     3893     +135     

Impacted Files	Coverage Δ
scopesim/optics/fov_utils.py	63.04% <100.00%> (+0.82%)	⬆️

... and 12 files with indirect coverage changes

☔ View full report in Codecov by Sentry.
📢 Do you have feedback about the report comment? Let us know in this issue.

[oczoske]

Sorted. The function needed protection against accessing pixels outside the ImageHDU.

[hugobuddel]

Thank you @oczoske. Tagging #209. I'll merge.

Maybe this missing protection against accessing pixels outside the ImageHDU was also the root cause of the MICADO notebooks failing for 4k*4k simulations back when you first tried to fix the off-by-one-error. Unfortunately I cannot now find the issue / PR where we discussed that.

That is, the original rounding code would usually round in the 'right' direction, but occasionally in the 'wrong' direction, which would only happen for simulations of large detectors. Your protection would probably fix that, so one step closer to doing this properly.