Fix for gp.sample_gp (redone)

[bwengals]

This is a quick patch for gp.sample_gp. It addresses the possible culprit of #2015, and the cause of #2137. The old PR is #2261.

I think some people prefer jitter, others prefer chol_const. Happy to change per suggestions.

[twiecki]

I think jitter we already use for a random term, this isn't really jittering but just adding a small constant, so I would prefer the chol_const name.

[bwengals]

sounds good, changed it

[twiecki]

______________________________ TestGP.test_sample ______________________________
self = <pymc3.tests.test_gp.TestGP object at 0x7f3c9231f250>
    def test_sample(self):
        X = np.linspace(0, 1, 10)[:, None]
        Y = np.random.randn(10)
        with Model() as model:
            M = gp.mean.Zero()
            l = Uniform('l', 0, 5)
            K = gp.cov.Matern32(1, l)
            sigma = Uniform('sigma', 0, 10)
            # make a Gaussian model
            random_test = gp.GP('random_test', mean_func=M, cov_func=K, sigma=sigma, observed={'X':X, 'Y':Y})
            tr = sample(20, init=None, progressbar=False, random_seed=self.random_seed)
    
        # test prediction
        Z = np.linspace(0, 1, 5)[:, None]
        with model:
            out = gp.sample_gp(tr[-3:], gp=random_test, X_values=Z, obs_noise=False,
>                              random_seed=self.random_seed, progressbar=False, jitter=True)
E           TypeError: sample_gp() got an unexpected keyword argument 'jitter'
pymc3/tests/test_gp.py:418: TypeError

[bwengals]

ah whoops, fixed