bfsave: use the metadata endianness to create the byte array

[sbesson]

See https://forum.image.sc/t/issues-with-bfsave-and-native-ome-metadata/106195

The existing code assumes the data are ordered using big-endian as in the default minimal OME-XML metadata.
This adjusts the code to handle OME metadata specifying little-endian ordering

Testing can be done by downloading the LSM file uploaded https://zenodo.org/records/14510432 at running the minimal code (https://matlab.mathworks.com/ might be

data=bfopen('10-01.lsm');
bfsave(cat(3,data{1}{:,1}),'10-01.ome.tiff','metadata',data{4});

Without this PR, the OME-TIFF should appear corrupted. With this PR included, the array ordering should match the metadata and the output of showinf 10-01.lsm should match showinf 10-01.ome.tiff.

Additional tests should probably be performed e.g. using synthetic sample files with different endianness and pixel types to ensure that the logic works in all scenarios.

[imagesc-bot]

This pull request has been mentioned on Image.sc Forum. There might be relevant details there:

https://forum.image.sc/t/issues-with-bfsave-and-native-ome-metadata/106195/4

[melissalinkert]

Tested on matlab.mathworks.com with synthetic data in several pixel types and both little/big endian:

>> data = bfopen('test&pixelType=float.fake')
Reading series #1
    .

data =

  1x4 cell array

    {1x2 cell}    {0 java.util.Hashtable}    {1x1 cell}    {1x1 loci.formats.ome.OMEPyramidStore}

>> bfsave(cat(3, data{1}{:1}), 'float-little.ome.tiff', 'metadata', data{4});
 bfsave(cat(3, data{1}{:1}), 'float-little.ome.tiff', 'metadata', data{4});
                        ↑
Invalid expression. Check for missing multiplication operator, missing or unbalanced delimiters, or other syntax error. To construct matrices, use brackets instead of parentheses.
 
>> bfsave(cat(3, data{1}{:,1}), 'float-little.ome.tiff', 'metadata', data{4});
>> data = bfopen('test&pixelType=float&little=false.fake')
Reading series #1
    .

data =

  1x4 cell array

    {1x2 cell}    {0 java.util.Hashtable}    {1x1 cell}    {1x1 loci.formats.ome.OMEPyramidStore}

>> bfsave(cat(3, data{1}{:,1}), 'float-big.ome.tiff', 'metadata', data{4});
>> data = bfopen('test&pixelType=int32.fake')
Reading series #1
    .

data =

  1x4 cell array

    {1x2 cell}    {0 java.util.Hashtable}    {1x1 cell}    {1x1 loci.formats.ome.OMEPyramidStore}

>> bfsave(cat(3, data{1}{:,1}), 'int32-little.ome.tiff', 'metadata', data{4});
>> data = bfopen('test&pixelType=int32&little=false.fake')
Reading series #1
    .

data =

  1x4 cell array

    {1x2 cell}    {0 java.util.Hashtable}    {1x1 cell}    {1x1 loci.formats.ome.OMEPyramidStore}

>> bfsave(cat(3, data{1}{:,1}), 'int32-big.ome.tiff', 'metadata', data{4});
>> data = bfopen('test&pixelType=uint16.fake')
Reading series #1
    .

data =

  1x4 cell array

    {1x2 cell}    {0 java.util.Hashtable}    {1x1 cell}    {1x1 loci.formats.ome.OMEPyramidStore}

>> bfsave(cat(3, data{1}{:,1}), 'uint16-little.ome.tiff', 'metadata', data{4});
>> data = bfopen('test&pixelType=uint16&little=false.fake')
Reading series #1
    .

data =

  1x4 cell array

    {1x2 cell}    {0 java.util.Hashtable}    {1x1 cell}    {1x1 loci.formats.ome.OMEPyramidStore}

>> bfsave(cat(3, data{1}{:,1}), 'uint16-big.ome.tiff', 'metadata', data{4});
>> data = bfopen('test&pixelType=uint8.fake')
Reading series #1
    .

data =

  1x4 cell array

    {1x2 cell}    {0 java.util.Hashtable}    {1x1 cell}    {1x1 loci.formats.ome.OMEPyramidStore}

>> bfsave(cat(3, data{1}{:,1}), 'uint8-little.ome.tiff', 'metadata', data{4});
>> data = bfopen('test&pixelType=uint8&little=false.fake')
Reading series #1
    .

data =

  1x4 [cell](matlab:helpPopup('cell')) array

    {1x2 cell}    {0 java.util.Hashtable}    {1x1 cell}    {1x1 loci.formats.ome.OMEPyramidStore}

>> bfsave(cat(3, data{1}{:,1}), 'uint8-big.ome.tiff', 'metadata', data{4});

Output .ome.tiff files all look as expected and have correct pixel type/endianness.

[melissalinkert]

Will plan to merge this tomorrow if builds continue to pass overnight.

[melissalinkert]

https://merge-ci.openmicroscopy.org/jenkins/job/BIOFORMATS-test-repo/296/ now passing with this included.