Consider redesign of MemoryReader.release and MemoryWriter.release

[CommanderTvis]

No description provided.

[altavir]

I agree, but I do not have any clear ideas about what to do with them. I think it will be more clear when we have some kind of prototype of working with native or memory-mapped blocks.

[CommanderTvis]

Well, current usages of Memory inside the library rely on that off-heap memory is managed by GC.

[CommanderTvis]

So, we have either to carefully use such objects like MemoryBuffer inside certain scopes or not to use really native memory without any hacks like Cleaner.

[CommanderTvis]

I agree, but I do not have any clear ideas about what to do with them. I think it will be more clear when we have some kind of prototype of working with native or memory-mapped blocks.

An obvious prototype of using managed native objects is kmath-gsl.

[altavir]

We need this interface to be usable both for managed and for non-managed memory. So there should be some kind of either release method or strict scoping policy. It does not make a lot of sense for only managed memory. I am not quite sure right now, but I think that there should be a release method like it is now, and the scoping rules should be built on top of it.

[CommanderTvis]

Maybe we should separate managed memory for things like MemoryBuffer from manual one?

[CommanderTvis]

Closing because of #432