DAS-2259 fix opera rtc browse images.

[flamingbear]

Description

A well placed, careful, cast to uint8, prevents trying to create PIL images from array with float32 values.

Also updates numpy to v2. because it has the np.min_scalar_type used in debugging.

Jira Issue ID

DAS-2259

Local Test Steps

• Check out this branch, build and run the tests.

❯ ./bin/build-image && ./bin/build-test && ./bin/run-test

• Verify all tests pass. Fire up Harmony-In-A-Box and run the regression test notebook against localhost.

• Examine the notebook explain-opera.pdf attached to the DAS-2259 ticket for explanation of the change.

• Lastly generate an image for OPERA RTC:

  1. Download the input tif on the DAS-2259 ticket
     OPERA_L2_RTC-S1_T039-082699-IW3_20240909T092236Z_20240909T134139Z_S1A_30_v1.0_rgb.tif

  2. Run the script below (after installing hybig as editable in your environment)

   from hybig import create_browse
   fn = './OPERA_L2_RTC-S1_T039-082699-IW3_20240909T092236Z_20240909T134139Z_S1A_30_v1.0_rgb.tif'
   create_browse(fn, {})

  3. bask in the glow of output PNGs.

PR Acceptance Checklist

• Jira ticket acceptance criteria met.
• CHANGELOG.md updated to include high level summary of PR changes.
• docker/service_version.txt updated if publishing a release.
• Tests added/updated and passing.
• Documentation updated (if needed).

[lyonthefrog]

All unit tests and HyBIG_Regression.ipynb tests passedand Ian the little test script and output the expected PNG:

It was a simple change, I don't have any more comments or questions, so I'll approve on my end.

[owenlittlejohns]

Nice work - looks good. I also got an image with a transparent region around the data. The unit tests ran and I successfully ran the regression tests against the image I built locally.

It took me a moment to grok the numbers in the expected test output (remembering that the values would be normalised in the band such that they extend from the minimum to the maximum value), but that was just a me thing getting back into the swing of HyBIG.

[flamingbear]

moment to grok the numbers in the expected test output

Yeah, I wasn't sure it was super valuable other than to show the dtypes would work correctly for just that reason. And I just grabbed values from an existing image so they might not be super elucidating.