Add in JNI for parsing JSON data and getting the metadata back too.

[revans2]

Description

Adds in a new java binding to allow reading a JSON buffer and getting back the metadata along with the table when inferring the schema.

Checklist

• I am familiar with the Contributing Guidelines.
• New or existing tests cover these changes. (the functionality is covered by C++ tests, this just adds in a new java binding, if more is needed/wanted I am happy to add more)
• The documentation is up to date with these changes.

[jbrennan333]

This looks good to me.

[codecov]

Codecov Report

❗ No coverage uploaded for pull request base (branch-22.10@039622f). Click here to learn what that means.
The diff coverage is n/a.

@@               Coverage Diff               @@
##             branch-22.10   #11431   +/-   ##
===============================================
  Coverage                ?   86.47%           
===============================================
  Files                   ?      144           
  Lines                   ?    22856           
  Branches                ?        0           
===============================================
  Hits                    ?    19764           
  Misses                  ?     3092           
  Partials                ?        0           

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[ttnghia]

Can this be implemented as a derived class extending Table?

[revans2]

It would not match what the C++ code is doing. If you want me to try I can. I agree that conceptually it would be more interesting, but I wanted to preserve the C++ API in this case. Also it gets to be a more invasive change to make it work.

[ttnghia]

I see. So it is just some wrapper for special purposes.

[revans2]

Yes. For all of the other APIs where we read the from a file we already know the names of the columns, as we passed them into the reader. So in those cases we just return a Table with the columns in the order that was requested when it was passed in. Here we need the metadata to be able to know what the names of the columns are. Long term we should probably update all of our reader APIs to match what CUDF is doing in C++ and just return the TableWithMeta. But for now to minimize the impact of the change we decided not to do that. I do like the idea of having TableWithMeta be a Table too. We probably will/should implement that if we do try to update all of the other APIs, just because it would minimize the impact of the change on customer code.

[ttnghia]

Some Python tests failed. Let's try to merge upstream.

[revans2]

@gpucibot merge