[BUG] Auglag algorithm from NLOPT library does not evolve the population

**Describe the bug**
I am trying to use `auglag` and `auglag_eq` algorithms from the NLOPT library. However, when I run the code the population does not evolve. In fact, I checked if the gradient was being evaluated, but it wasn't. So, the model is running, but it's not evolving.

It is important to note that all the other algorithms available for pygmo are working correctly.

**To Reproduce**
The code is quite large, so I will summarise here the structure:

`class pygmo_run:`
`prob = pg.problem(my_udp)`
`nl_algo = pg.nlopt('auglag')`
`nl_algo.local_optimizer = pg.nlopt('slsqp')`
`algo = pg.algorithm(uda = nl_algo)`
`pop = pg.population(prob, size=10, seed = 0)`
`pop = algo.evolve(pop)`

`class my_udp:`
`def fitness:`
`def get_bounds:`
`def gradient:`

**Expected behavior**
The population should evolve and search for the minimum.

**Environment (please complete the following information):**
- OS: Windows 10
- Installation method: conda
- Version: 2.19

Start by reproducing the reported behavior with pg.nlopt('auglag'), the SLSQP local optimizer, and the my_udp structure described in the issue. Compare gradient evaluation and population results with a working algorithm; done means auglag and auglag_eq evolve the population and search for the minimum, ideally with a small complete reproducer and a regression test.

Written by the indexing model from the issue text.