`lispyville-end-of-defun' doesn't go to the next defun when point is
already at the end of a defun, whereas `lispyville-beginning-of-defun'
does. This commit fixes this inconsistency by matching
`lispyville-end-of-defun''s behaviour with `lispyville-beginning-of-defun'.
`line-end-position' isn't enough because it never returns `t' in normal
mode (in normal mode the point can't reach the true end of a line).