Neural Implicit Action Fields: From Discrete Waypoints to Continuous Functions for Vision-Language-Action Models