Apply rtos_fix hack for handling 32-bit floats on all calls to rtos_fix

[aaronfranke]

This PR was made in response to this comment #98750 (comment)

I don't think there is any harm in always applying this hack to all float serialization. As a brief reminder, in all cases, this is strictly better than the behavior in Godot 4.4.

In Godot 4.4 and earlier, it would always serialize 6 digits of a float, which resulted in precision loss when serializing. In Godot 4.5, it uses Grisu2 instead to precisely serialize, however this can cause problems when Godot tells Grisu2 to serialize a double when the underlying data was a 32-bit float. I added a hack in PR #98750 that checks if it's equivalent to the 32-bit float value, and then serializes assuming the value is a 32-bit float, which seamlessly avoids issues for users, though it does mean that for double sources, 1 in 2^29 values will be off by 2^-23.

This PR changes this to apply to all floats, including Vector2, Vector3, Packed*Arrays, etc. In practice, this only changes behavior for Vector2, Vector3, etc for double-precision builds of Godot, and it has just one behavior change for single-precision builds of Godot, PackedFloat64Array. Test code:

func _ready() -> void:
	var vec := Vector2(0.1, 0.2)
	print(var_to_str(vec))
	print(var_to_str(PackedFloat64Array([vec.x, vec.y])))

In master, this is printed:

Vector2(0.1, 0.2)
PackedFloat64Array(0.10000000149011612, 0.20000000298023224)

With this PR, this is printed:

Vector2(0.1, 0.2)
PackedFloat64Array(0.1, 0.2)

[cridenour]

Tested and working as described. A huge QoL upgrade for those on precision=double. Thank you @aaronfranke

[Ivorforce]

Makes sense to me.

[Calinou]

Tested locally, it works as expected. Code looks good to me.

When using Tools > Upgrade Project Files on the 3D platformer demo, this is the only change:

 [node name="Circle" type="MeshInstance3D" parent="."]
-transform = Transform3D(1, 0, 0, 0, -3.61999e-06, -1, 0, 1, -3.61999e-06, 0.00550658, 0.532125, 0.000651598)
+transform = Transform3D(1, 0, 0, 0, -3.6199901e-06, -1, 0, 1, -3.6199901e-06, 0.00550658, 0.532125, 0.000651598)

I've tried various other demos and changes are similarly minimal, so this change should be safe in terms of VCS diffs.

[Repiteo]

Thanks!