VM-alloc branch-and-bound

Improved class hierarchy.

Added base class for allocator parameters and derived the BnB and ILP parameter classes from here.

VMAllocator is now an interface. The BnB and ILP allocators implement this interface.

Updated the config parser to handle the polymorphic parameter classes.

ILP solvers are configured in the config file similarly to BnB solvers.

Updated the sample config file.

Evaluation improvements and fixes.

- Changed active hosts coefficient to 10
- Added initial lower bound computation
- Migration limit is now a ratio
- Tests for multiple sizes in steps available
- Input problem data written to csv
- Cost component data written to csv
- lp_solve path can be configured
- Fixed crash when gurobi couldn't find a result
- Limited gurobi to 1 thread
- Fixed a VM sorting bug

Intelligent bound now available.

If enabled, the allocator solves a relaxed problem after every allocation.

This is done in order to determine a more accurate lower bound for the complete allocation.

Minor changes:
- Added intelligentBound parameter to configParser
- Problem instance ID is now logged in main.cpp
- Introduced additional VM counter in PM.h, required for bounding
- Updated TODO

Hotfix for PM comparators.

Fixed PM order when one PM is off and the other one is on (this was wrong in all comparators).

Fixed the case where both PMs are off for the sum comparator.

Improved PM sorting.

PM sorting now differentiates between PMs which are on/off.
-> PMs on are sorted in ascending order and put at the start of the list.
-> PMs off are sorted in descending order and put at the end of the list.

Also fixed a typo in the SortType enum.

Added simplistic configuration file support.

A sample config file is added for demonstration purposes.

Other minor modifications:
- Added toString function for SortType
- Hotfix for unsafe function sprintf
- Added WIN32 defines for ILP solvers
- Updated makefile
- Added header guard to ProblemGenerator.h
- Updated VS project

ProblemGenerator can now do a First-Fit allocation so that the generated initial mapping is not that bad

Added wrapper to ILP solvers Gurobi and Lp_solve

Other sorting methods for VMs, VMs without an initial PM.

Added different sorting methods for VMs, similarly to the way it works by PMs.

VMs without an initial PM are now supported, allocating these never cause migrations, regardless of the destination.

Migration limit, iterative cost computation.

Introduced a new parameter with which the number of migrations allowed can be limited.

The cost of an allocation is now computed iteratively, instead of recalculating the total cost at each allocation. The number of PMs turned on and the amount of migrations are tracked in two member variables.