Fix #180: Improve efficiency of burn-in phase in sampling functions

[Konikz]

Description

This PR improves the efficiency of the burn-in phase in sampling functions by avoiding unnecessary storage and deletion of burn-in points. Instead of storing points during burn-in and then clearing them, we now use a NoOpWalkPolicy that simply discards these points.

Changes

• Added NoOpWalkPolicy that discards points during burn-in phase
• Modified sampling functions to use NoOpWalkPolicy instead of store-and-clear approach
• Added performance test to verify improvements

Before

if (nburns > 0) {
    RandomPointGenerator::apply(P, p, nburns, walk_len, randPoints,
                                push_back_policy, rng);
    randPoints.clear();
}

After

if (nburns > 0) {
    RandomPointGenerator::apply(P, p, nburns, walk_len, randPoints,
                                no_op_policy, rng);
}

Benefits

• Eliminates unnecessary memory allocations
• Reduces memory usage during burn-in phase
• Maintains same functionality and API compatibility

Testing

Added burn_in_efficiency_test.cpp that compares performance and verifies correctness of both implementations.

Fixes #180