OMERO_Matlab:Fix the casting from *unit8 to int8

[bramalingam]

What this PR does

ref:
https://www.openmicroscopy.org/community/viewtopic.php?f=6&t=8117

Testing this PR

1. Test the Matlab toolbox with the code snippet given in the forum post.
2. Check if the test-code runs without any errors.

[sbesson]

I wonder if it would make sense here to take advantage of the minimal example provided by the user in the forum post and extend the WriteData training example to upload a mask created with imread, retrieve it and compare both input and output are equal?

[bramalingam]

Yes that makes sense, will add that minimal test to the writeData training example. Although the downloaded version comparison cannot happen as suggested by the user. (unless we want to check the downloads aspect as well in this test)

[jburel]

This is a candidate PR for 5.2.6. see https://trello.com/c/oWtA5Pod/9-omero-matlab-regression#

[imunro]

PR Looks Good to me. FYI I've been doing some testing locally and it looks as if any string starting with '*' is invalid and therefore Matlab (Silently) uses the default which is uint8. Writing test data to a file it seems as if 0-7F hex are written fine but anything outside this range is written , to the file, as 00. The results will be unpredictable depending on the original data type.

[jburel]

@bramalingam could you rebase this PR to dev_5_2?

[bramalingam]

--rebased-to #4849

[jburel]

@imunro's comment: see card https://trello.com/c/Wq0xVNHc/191-matlab-invalid-string