Added bmrmopt class in shogun/optimization/framework

[ghost]

@vigsterkr
The previous PR got corrupted for some reason.
This PR corresponds to issue #1969
bmrmopt is a simple class derived from libbmrm.cpp/h. This is the first class in shogun/optimization/framwork. Other bundle solvers can be easily derived from this class.
A brief overview of the files:
1.) functions.h/cpp - func class for handling functions and solution vectors
2.) constraints.h/cpp - constraints class for handling constraints. (For now only simplex constraints)
3.) CP.h/cpp - contains class for handling cutting planes
4.) optdefines.h - general include file
5.) bmrmopt.h/cpp - contains bmrmopt class (libbmrm.cpp). Other classes can be easily derived from this class. Will add them shortly in the same directory. (ppbmopt, p3bmopt, ncbmopt)

[ghost]

@vigsterkr Can you please look it? Build is already successful.

[vigsterkr]

@kprah honestly i don't know how to write a comment on this that i don't feel that i'm being rude. I know that you are working hard to try to solve this problem, and this is your 3rd PR, which i really appreciate, but.....
but as the last 2 times i've told you it would have been much better to actually do some planning, i.e. have some discussion about how and what to do and then execute it. You've told me that you gonna do this this time, but again you've disappeared for couple of days and now you came back with this PR....

about the PR itself:
my main concern is not that you are not following any of Shogun's code and design patterns (which already a problem), but you don't follow any object oriented design pattern, which is very essential when you use C++ for coding. for example in all your classes everything is defined public, you create classes not for representing entities, but for representing functions...

i'm sorry if i offended you in any way, i'm really trying to do some constructive criticism here...

[lisitsyn]

It makes sense to use initializer list here