Gpytorch: Initializing GaussianLikelihood noise broken

[KeAWang]

>>> import gpytorch
>>> gl = gpytorch.likelihoods.GaussianLikelihood()
>>> gl.initialize(noise=1)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File ".../gpytorch/gpytorch/module.py", line 89, in initialize
    setattr(self, name, val)
  File "../lib/python3.6/site-packages/torch/nn/modules/module.py", line 579, in __setattr__
    object.__setattr__(self, name, value)
  File ".../gpytorch/gpytorch/likelihoods/gaussian_likelihood.py", line 63, in noise
    self.noise_covar.initialize(value)
TypeError: initialize() takes 1 positional argument but 2 were given

raw_noise works though

>>> gl.initialize(raw_noise=1)
GaussianLikelihood(
  (noise_covar): HomoskedasticNoise()
)

[gpleiss]

We're open to a PR 😄

[Balandat]

This might be a good opportunity to remove the log_ kwargs deprecation... this has been around for a while now and we could declutter things...