Take the old-fashioned approach and expand the solution in a polynomial in eccentricity.
Stick this back into the Kepler equation, then expand on eccentricity to get
Solve successively for the coefficients. We'll write a quickie procedure to do this automatically.
Check:
Hence, the approximate solution is
Plot just the coefficients.
plot
Plot the difference between the approximate solution and the mean anomaly M (which is the zeroth-order approximation), as a function of eccentricity and M.
plot
Plot only the highest-order term of the approximate solution.
plot
Create a function to numerically solve for the eccentric anomaly, then plot the error of the approximate solution.
plot