Parallel version of random walks

[Kostas-Pallikaris]

This PR provides multi-threaded version for the following random walks:

• Random Directions Hit-and-Run (uniform and Gaussian sampling)
• Coordinate Directions Hit-and-Run (uniform and Gaussian sampling)
• Billiard walk
• Random Directions Hit-and-Run for boundary sampling
• Coordinate Directions Hit-and-Run for boundary sampling

The interface is written in C++ and uses the library openmp.

[vissarion]

👋 @Kostas-Pallikaris what is the status of this PR? Is it ready for review?

[codecov]

Codecov Report

Merging #172 (a1d47e6) into develop (6591ae8) will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff            @@
##           develop     #172   +/-   ##
========================================
  Coverage    55.44%   55.44%           
========================================
  Files           85       85           
  Lines         4801     4801           
  Branches      2111     2111           
========================================
  Hits          2662     2662           
  Misses         889      889           
  Partials      1250     1250           

Impacted Files	Coverage Δ
include/random_walks/boundary_rdhr_walk.hpp	68.00% <ø> (ø)
include/random_walks/uniform_cdhr_walk.hpp	92.85% <ø> (ø)

[TolisChal]

@Kostas-Pallikaris, thanks for the revised version!

[vissarion]

Thanks for this PR!

I have some comments.

I would also suggest to include all walks in one file in include/random_walks e.g. multithread_walks.hpp.

It would be very useful to add an example that uses the new functionality as a use case.

[vissarion]

What about this

std::vector<int> num_points_per_thread(rnum%num_threads, rnum/num_threads+1);
std::vector<int> a(num_threads - rnum%num_threads, rnum/num_threads);
num_points_per_thread.insert(num_points_per_thread.end(), a.begin(), a.end());

instead of the double loop? Seems simpler and more efficient.

[Kostas-Pallikaris]

Thank you @vissarion. I've implemented this.

[vissarion]

I am ok with merging now. Thanks!
I have one more minor comment

[vissarion]

you do not have to repeat it here, just inherit the method from policy_storing<BRDHRWalk_multithread> i.e.

template <>
struct policy_storing<BCDHRWalk_multithread> : policy_storing<BRDHRWalk_multithread> 
{}

[Kostas-Pallikaris]

Thank you @vissarion. I've implemented this change.