Perf issue: avoid vector copy at each insertion.

[greg7mdp]

The vector::reserve(vector::size() + 1) call which is updated in this PR, with typical std::vector implementations, causes the vector to grow by a single element on every reverve, and therefore a memory allocation and full vector copy on every reserve call.

This updates the reserve() call so that the vector capacity grows quadrically.

[ericpassmore]

Note:start
group: STABILITY
category: PERFORMANCE
summary: Updates the reserve() call so that the vector capacity grows slower at a quadrically rate.
Note:end