Expose the rest of the Driver API

[bheisler]

RustaCUDA currently exposes only a minimum-viable subset of the Driver API. Ultimately, it should expose all non-deprecated public functions in the API.

This is a tracking issue to keep track of which aspects of the API should be available but aren't. If you need something that isn't listed here, add a comment and raise an issue.

• Events: Add support for Events #6
• Primary Context Management: Add support for Primary Context Management #7
• Asynchronous memcpy: Add support for asynchronous memcpy #8
• JIT Linking: Add support for JIT linking #9
• Multi-dimensional arrays: Add support for CUDA arrays #10
• Texture/surface objects: Add support for textures and surfaces #11
• Expose the rest of the Unified memory management API: Add complete support for Unified Memory #12
• Stream callbacks: Expose Stream Callbacks #13

There are more features to add (including features only added after CUDA 8.0) but this should be enough for now.

[phdoerfler]

@bheisler I'm still fairly new to CUDA so please excuse me if this makes little sense: I'd love to see support for using tensor cores in RustaCUDA. My current project requires shared memory (which I believe RustaCUDA supports?) and tensor cores for which I found no sign that they are supported. If I can use those with RustaCUDA I'll switch in a heartbeat.