Applied master branch bug fixes.

- Planner was under-estimating maximum speeds through straight
junctions in certain cases. The calculations have been updated to be
more accurate.

- Strange sizeof() bug in the most recent releases. Manifested as an
alarm upon a power up even when homing was disabled. Fixed by declaring
sizeof() with struct types, rather than variable names, even though
they were validated to give the same value.

- Spindle speed zero should disable the spindle. Now fixed.

- New configuration option for inverting certain limit pins. Handy for
mixed NO and NC switch machines. See config.h for details.

grbl/planner.c

@@ -200,7 +200,7 @@ static void planner_recalculate()
 
 void plan_reset() 
 {
-  memset(&pl, 0, sizeof(pl)); // Clear planner struct
+  memset(&pl, 0, sizeof(planner_t)); // Clear planner struct
   block_buffer_tail = 0;
   block_buffer_head = 0; // Empty = tail
   next_buffer_head = 1; // plan_next_block_index(block_buffer_head)
@@ -394,11 +394,11 @@ uint8_t plan_check_full_buffer()
        change the overall maximum entry speed conditions of all blocks.
     */
     // NOTE: Computed without any expensive trig, sin() or acos(), by trig half angle identity of cos(theta).
-    if (junction_cos_theta > 0.99) {
+    if (junction_cos_theta > 0.999999) {
       //  For a 0 degree acute junction, just set minimum junction speed. 
       block->max_junction_speed_sqr = MINIMUM_JUNCTION_SPEED*MINIMUM_JUNCTION_SPEED;
     } else {
-      junction_cos_theta = max(junction_cos_theta,-0.99); // Check for numerical round-off to avoid divide by zero.
+      junction_cos_theta = max(junction_cos_theta,-0.999999); // Check for numerical round-off to avoid divide by zero.
       float sin_theta_d2 = sqrt(0.5*(1.0-junction_cos_theta)); // Trig half angle identity. Always positive.
 
       // TODO: Technically, the acceleration used in calculation needs to be limited by the minimum of the