Bounded stack

[lyrm]

Based on #155. This PR removes the blocking part of the previous PR. The issue with it is that it adds a dependency to picos, probably requiring to create a Saturn_picos package.

For Saturn.1.0, we have decided to do only a bounded stack, thus this PR.

[lyrm]

@Sudha247 : Could you have a look at this PR (at the documentation, or even a quick look at the tests, maybe ?), any feedbacks will be great !

[art-w]

Looking good! Some minor comments

[lyrm]

@art-w I think this PR is good to. What do you think ?

[art-w]

Yes it looks good! Out of curiosity, are the copy_as_padded on records necessary? (my understanding is that this makes the allocated object fatter, so as to avoid false-sharing contention, but the head atomic already appears to be protected by make_contended)

[lyrm]

Mmh there is a stack overflow on Treiber_stack.push_all. This PR has almost the same implementation so I will have a look before merging !

[lyrm]

Yes it looks good! Out of curiosity, are the copy_as_padded on records necessary? (my understanding is that this makes the allocated object fatter, so as to avoid false-sharing contention, but the head atomic already appears to be protected by make_contended)

Good question. As nothing else is mutable in the stack type, probably not.