Ability to create float16 and unsigned integer tensors using environment

[nk2IsHere]

Is your feature request related to a problem? Please describe.

We would like to be able to create float16 and unsigned integer tensors using newTensor function. Currently attempting to make such tensors results in UnsupportedOperationException "OnnxTensor with type ... is not supported".

Describe the solution you'd like

N/A

Describe alternatives you've considered

N/A

Additional context

We'd like to implement it and submit a PR.

[yuzawa-san]

please review #421

the pr you sent i think does not handle it correctly. it uses a float buffer, but in reality the elements are 16 bits, so a short buffer should be used. i tried to align everything properly in my version and also i added support for all types the library exposes.

[yuzawa-san]

fixed in #421