Remove duplicated calculation of coordinates.

[shishaochen]

In function quasi_newton::operator(), the model coordinates are always updated with the conformer searched from BFGS optimization.

AutoDock-Vina/src/lib/quasi_newton.cpp

Lines 43 to 51 in 350bb7f

	void quasi_newton::operator()(model& m, const precalculate_byatom& p, const igrid& ig, output_type& out, change& g, const vec& v, int& evalcount) const { // g must have correct size
	quasi_newton_aux aux(&m, &p, &ig, v);
	fl res = bfgs(aux, out.c, g, max_steps, average_required_improvement, 10, evalcount);
	// Update model a last time after optimization
	m.set(out.c);
	out.e = res;
	}

Thus, there is no need to calculate model coordinates in function monte_carlo::operator().

AutoDock-Vina/src/lib/monte_carlo.cpp

Lines 59 to 74 in 350bb7f

	quasi_newton_par(m, p, ig, candidate, g, hunt_cap, evalcount);
	if(step == 0 || metropolis_accept(tmp.e, candidate.e, temperature, generator)) {
	tmp = candidate;
	m.set(tmp.c); // FIXME? useless?
	// FIXME only for very promising ones
	if(tmp.e < best_e || out.size() < num_saved_mins) {
	quasi_newton_par(m, p, ig, tmp, g, authentic_v, evalcount);
	m.set(tmp.c); // FIXME? useless?
	tmp.coords = m.get_heavy_atom_movable_coords();
	add_to_output_container(out, tmp, min_rmsd, num_saved_mins); // 20 - max size
	if(tmp.e < best_e)
	best_e = tmp.e;
	}
	}

To ensure consistency, a test is executed between the base commit and the updated one. Bash commands are pasted below:

cd example/basic_docking/
time vina \
  --receptor solution/1iep_receptor.pdbqt \
  --ligand solution/1iep_ligand.pdbqt \
  --config solution/1iep_receptor_vina_box.txt \
  --exhaustiveness 16 \
  --seed 137 \
  --cpu 6 \
  --no_refine  # Observe the output of Monte-Carlo search

It can be verified that generated files solution/1iep_ligand_out.pdbqt are exactly alike.

With this pull request merged, we are convinced that CPU cost will be reduced slightly. My CPU type is Intel(R) Xeon(R) Platinum 8336C CPU @ 2.30GHz and average time of 10 runs are:

version	real	user	sys
baseline	33.83s	180.47s	0.37s
updated	33.76s	180.20s	0.35s