TST: stats.rdist: skip failing xslow test

[mdhaber]

Reference issue

Closes gh-21949

What does this implement/fix?

gh-21949 reported an xslow fit test failure. We changed from passing an integer seed to differential_evolution via rng instead of random_state, so it makes sense that we'd get some change in this test. We already adjusted tests after that (gh-21822), so this is something different.

Additional information

This generic test is performed for all distributions, but since our optimization techniques do not guarantee convergence to the global optimum for all possible problems, some tests fail. Sometimes we'll skip the generic test for a distribution and replace it with another one that is slightly easier for the optimizer to solve, e.g.

import matplotlib.pyplot as plt
import numpy as np
from scipy import stats
from scipy.optimize import differential_evolution as de
from scipy.stats.tests.test_fit import assert_nlff_less_or_close

N = 1000
rng = np.random.default_rng(23854582452784625)
dist = stats.rdist
shapes = (1., 2., 3.)
data = dist.rvs(*shapes, size=N, random_state=rng)
shape_bounds = {'loc': (0.1, 10), 'scale': (0.1, 10), 'c': (0.1, 10)}
res = stats.fit(dist, data, shape_bounds, method='mse',
                optimizer=lambda *args, **kwds: de(*args, rng=rng, **kwds))
assert_nlff_less_or_close(dist, data, res.params, shapes, nlff_name='_penalized_nlpsf')
res.plot()
plt.show()

In this case, I think it's safe to just skip the test. beta has always had trouble, so it's not so surprising that rdist (symmetric beta) has trouble, and beta has a distribution-specific test, so I don't know that we need a separate one for a special case distribution.