Add support for specifying measurement uncertainty degrees of freedom/t-distribution

[mestinso]

As is mentioned here: #33 (comment), one feature potentially worth adding to this package is a way to specify the degrees of freedom for each measurement. As is, the package assumes a normal distribution where the standard deviation is known. In general, the standard deviation for a given measurement is not known, and instead, we may only have a sample standard deviation with an associated degrees of freedom, which follows a t distribution. The specified input degrees of freedom should be propagated to the output degrees of freedom (analogous to propagating uncertainty). I would propose a third argument could be added to the measurement function to specify the degrees of freedom. If unspecified, the degrees of freedom would be set to infinity (in line with the current assumption). To compliment this, I would suggest a helper function that estimates the expanded uncertainty at a user specified level of confidence.

[mestinso]

@giordano Any thoughts on this one?