With the data you linked to: http://spacemath.gsfc.nasa.gov/earth/6Page58.pdf
A linear fit seems fairly reasonable for the given data, but using it to predict the near future is not really possible, since the data deviates from a linear trend a fair amount.
I used linear regression to get that the change in the number of days in a year versus the number of years from now (in millions) is given by $$\Delta D = -0.13\Delta T$$
So if we let $\Delta D = -0.2425$, we get that $\Delta T = 1.87$. So we estimate it would take 1.87 million years to have exactly 365 days in a year, which is on the same order as the estimation made by babou (since we both assumed linearity).
However, as I said above, the data does deviate from the linear regression a fair amount. I found that the mean absolute difference between the linear regression and the data points was about 8 days (max = 19 days, min = 0.4 days), which is a lot more than the 0.2425 days that we are dealing with.