fix #11202: introduce suggestChecksumAlgorithm method to FS API

[mtbc]

The main thing to test is that this breaks nothing: imports still work, and SHA1 hashes are still properly noted in the originalfile DB table. Also, check that the ManagedRepositoryITest unit tests all pass.

--no-rebase as dev_4_4 is limited to only SHA1-160.
Fixes http://trac.openmicroscopy.org.uk/ome/ticket/11202.

[bpindelski]

I'd argue that Adler-32 should be at the last position, as it's popularity is low and it's the weakest (yet fastest) in sense of collisions.

[bpindelski]

Besides the tiny comment, unit tests work ok, import through CLI and Insight too. Looks good to merge.

[jburel]

@mtbc: no longer possible to import with this PR w/o insight specifying a checksum algorithm

[mtbc]

Yes: this is why I wanted to get the algorithm specification in earlier (i.e. into 5.0beta2 Insight) rather than later when we have more adopters. The point is to prevent the possibility that the server chooses a checksum algorithm that the client doesn't actually support, especially as the process doesn't complete without the client providing the server the correct hashes. (Also, to prevent the opposite situation.)

Then if, at some later point, what the server supports and what the client's implementation of ChecksumProviderFactory supports don't match too well, we can adjust the code to cope with that without any further API changes.

[mtbc]

Filed http://trac.openmicroscopy.org.uk/ome/ticket/11577 for the less urgent (not-API-breaking) work to finish this change off.

[mtbc]

@bpindelski's testing remains valid: I'll revert the revert at some later point!

[joshmoore]

@jburel : any other issues from insight you're aware of? or all covered under 11577? (And if so, for beta2?)

[jburel]

All "upload" will need to be carefully tested, can be done for beta2 just need to have a clear list. (I can prepare and check)

[joshmoore]

👍