[Fix] Use default tolerances for np.testing.allclose() in WPLI ground truth test

[Moritz-Alexander-Kern]

This PR addresses an issue when testing Elephant on JUSUF and Galileo100, see #569 .

For the ground truth test in WPLI, see:

elephant/elephant/test/test_phase_analysis.py

Lines 494 to 514 in 5b464ec

	with self.subTest(msg="Quantity input"):
	freq, wpli = elephant.phase_analysis.weighted_phase_lag_index(
	self.lfps1_real, self.lfps2_real, self.sf1_real)
	np.testing.assert_allclose(
	wpli, self.wpli_ground_truth_ft_connectivity_wpli_real,
	atol=self.tolerance, rtol=self.tolerance, equal_nan=True)
	# np.array-input
	with self.subTest(msg="np.array input"):
	freq, wpli = elephant.phase_analysis.weighted_phase_lag_index(
	self.lfps1_real.magnitude, self.lfps2_real.magnitude,
	self.sf1_real)
	np.testing.assert_allclose(
	wpli, self.wpli_ground_truth_ft_connectivity_wpli_real,
	atol=self.tolerance, rtol=self.tolerance, equal_nan=True)
	# neo.AnalogSignal-input
	with self.subTest(msg="neo.AnalogSignal input"):
	freq, wpli = elephant.phase_analysis.weighted_phase_lag_index(
	self.lfps1_real_AnalogSignal, self.lfps2_real_AnalogSignal)
	np.testing.assert_allclose(
	wpli, self.wpli_ground_truth_ft_connectivity_wpli_real,
	atol=self.tolerance, rtol=self.tolerance, equal_nan=True)

The default tolerance values were used for np.testing.allclose to prevent test failures in cases where the deviations very small and therefore negligible.

[coveralls]

coverage: 86.97%. remained the same when pulling de0d108 on INM-6:fix/WPLI_unit_test_tolerance into 844eb82 on NeuralEnsemble:master.

[Moritz-Alexander-Kern]

build wheels

[Moritz-Alexander-Kern]

Note: Review was done by @ojoenlanuca .