motion: Fix state tag truncation and replace float with double

[BsAtHome]

The state tag is used in communication between interpreter and motion. Data put into it usually comes from double sized types and got truncated to float size. Afterwards, the values would, for example, be moved to HAL pins and the truncated value would be extended to double again. This is obviously a bad plan.

This PR makes the state tag field_float of double type.

[rodw-au]

Need to take care here. What about changes to emcmot_hal_data->interp_feedrate?
Its defined as hal_float_t *interp_feedrate;
Not to mention my pending ones in #3995

    /* New Geometric Metadata Pins */
    hal_float_t *interp_arc_radius;
    hal_float_t *interp_arc_center_x;
    hal_float_t *interp_arc_center_y;
    hal_float_t *interp_arc_center_z;
    hal_float_t *interp_straight_heading;
    hal_float_t *interp_normal_heading;  
    hal_bit_t   *iscircle;

Which leads me to suggest another change to resolve the active_tags mess this touches which I was leaving unitil #3995 was committed.
add an extra float/double array to state_tags
double fields_extra_float[EX_FIELD_MAX_FIELDS]
Move all new pins in #3995 to this field and lock fields_float to the original fields (up to GM_FIELD_FLOAT_NAIVE_CAM_TOLERANCE) with an assert statement and apropiate comments in state_tags.h

[BsAtHome]

The name hal_float_t is a complete misnomer and not a C-type float. The underlying type is actually a double.

The name "float" is used many places in linuxCNC, but most float types were promoted to double because you need the precision in the calculation. The variable names and defined did not change, only the underlying type.