Add missing flag when encode_variant writes math types with doubles

[Zylann]

Fixes #57935

The remote inspector is based on encode_variant and decode_variant to send and receive data. decode_variant uses a flag within the type to know if math types in the data are using doubles, but encode_variant is always writing real_t, without specifying the flag. So I made it so the flag is set when the instance of Godot writing the data is compiled with doubles.

cc @aaronfranke

[aaronfranke]

I don't know if this is correct. The flag should be set when encoding Variant::VECTOR2 with a double-precision build of Godot, but what about some of the other cases? We don't want Variant::VECTOR2I or Variant::COLOR to have this flag set.

[aaronfranke]

Does this work? Untested. Also, PACKED_INT64_ARRAY and PACKED_FLOAT64_ARRAY should be hard-coded to always have this set.

[Zylann]

We don't want Variant::VECTOR2I or Variant::COLOR to have this flag set.

True, I wasn't actually sure about that. But when I checked, this flag is actually only used on floating-point math types, not on integer ones.
At the very least, maybe I can add the following to make it explicit?

		case Variant::VECTOR2:
		case Variant::VECTOR3:
		case Variant::PACKED_VECTOR2_ARRAY:
		case Variant::PACKED_VECTOR3_ARRAY:
		case Variant::QUATERNION:
		case Variant::PLANE:
		case Variant::TRANSFORM2D:
		case Variant::TRANSFORM3D:
		case Variant::BASIS:
		case Variant::RECT2:
		case Variant::AABB: {

Edit: ok you wrote the same suggestion :D

[aaronfranke]

I guess the PACKED_*64_ARRAYs don't need it. To me it would still make sense to include it anyway since it 64-bit arrays are indeed encoded as 64-bit, but I guess it doesn't matter.

[aaronfranke]

Tested and it works as expected. The code looks good.

[akien-mga]

Thanks!