Fixes a code comment in Concurrent::{Hash,Array}

[fxn]

This patch is actually a question.

The reason Hash can be chosen in MRI as implementation for Concurrent::Hash cannot be that the interpreter does not run threads in parallel, because if you issue a delete call, say, and there is a context switch, even if only one thread runs at a time, the internal state of the hash could be left inconsistent.

What we need for Hash to work, really, is that their functions are atomic in practice.

Not realeasing the GIL while C runs is something that I've heard, but I don't really know if my proposal is correct. Someone knowledgeable should validate it.

I could do the same in Concurrent::Array and others if there are more.

[pitr-ch]

@fxn yeah, your new comment is more accurate and correct. Thanks. Please update it for array as well.

[fxn]

Excellent, I have updated the analogous comment in Concurrent::Array.

I have noticed Concurrent::Set has a similar comment, but Set is partially written in Ruby. Therefore, I am not really sure Concurrent::Set is actually thread-safe in CRuby. (But if that is correct, would probably be out of the scope of this patch anyway.)

[thedarkone]

@fxn's wording is better.

[pitr-ch]

Thanks!