argument name inconsistency in setting jitter value with gpytorch

[LuhuanWu]

• Change the init argument of class _dtype_value_context.

  • Currently the init function for _dtype_value_context in linear_operator/settings.py is:
    def __init__(self, float=None, double=None, half=None):

  • This is not consistent with the init function for _dtype_value_context in gpytorch/settings.py, which is:
    def __init__(self, float_value=None, double_value=None, half_value=None):

  • I would suggest to unify the argument names, for example, make the both init functions as:
    def __init__(self, float_value=None, double_value=None, half_value=None):

[Balandat]

That makes sense to me. Will put up a PR

[gpleiss]

Should be closed by #47