Extract reward before observation (C++ and Python)

[gasse]

See #209

[gasse]

I reordered reward, observation everywhere to be consistent (even in the constructor where it does not matter).

The tests seem to pass, at least on my machine.

@AntoinePrv can you check the code for C++ (great)ness ?

[gasse]

All CI tests pass now. Good to go ?

[AntoinePrv]

There is a long commit history that does not seem to belong to this PR.

[gasse]

Oh no, I though I had merged that one before the release :O

@AntoinePrv that's because I have rebased dev onto master. I've rebased this PR as well, should look better now.

[gasse]

I guess v0.7.0 will have a very short life...

[gasse]

Thanks for the code review ! I'll make the changes ASAP.

[gasse]

I removed the auto&& universal pointer stuff, switched back to std::move() in returns, and changed the lambda for a ternary operator. I also put some stuff inside a private function, to avoid a bit of code duplication.

@dchetelat @AntoinePrv let me know how that looks to you now.

[gasse]

Ah it looks like some conflicting commits have not been handled correctly when I rebased dev onto master for v0.7.0 ... Thanks @dchetelat for noticing that !

[dchetelat]

Yeah I feel it's the third time we fix that lol. Rebasing chaos...
Also I rewrote again the ternary thing, the compiler and the code check looks happy with it. All good to go for me as well!