Add 16-bits TGA support

[Ithamar]

This adds support for TGA files with 16 bits per pixel (basically, RGBA5551, where the alpha bit is optional).
It should be possible to cherry-pick this into 3.x branch with no conflicts too.

Closes #65715

[fire]

We made a decision not to support dithered artwork, but not sure the current state of this decision.

[Ithamar]

We made a decision not to support dithered artwork, but not sure the current state of this decision.

Does that also mean that the RGB565 / RGB5551 Image formats would be dropped in 4.x?

[Calinou]

We made a decision not to support dithered artwork, but not sure the current state of this decision.

This is only for import – the resulting image format remains whatever Godot supports.

Does that also mean that the RGB565 / RGB5551 Image formats would be dropped in 4.x?

No, they won't be dropped. I think @fire is referring to paletted (256-color) textures, which Godot can import (e.g. from PNGs) but GPUs haven't supported for decades. These are converted to RGB8 or RGBA8 on import depending on whether the PNG contains an alpha channel.

[LauPhi]

I see that the RGBA5551 TGA is still being saved as an Image::FORMAT_RGBA8 (32-bit).
Would it be possible to save it as a proper Image::FORMAT_RGBA5551 (16-bit) so that it would take two times less VRAM while loaded in the GPU?

[Ithamar]

I see that the RGBA5551 TGA is still being saved as an Image::FORMAT_RGBA8 (32-bit). Would it be possible to save it as a proper Image::FORMAT_RGBA5551 (16-bit) so that it would take two times less VRAM while loaded in the GPU?

It seems the TGA loader simply converts anything it reads to FORMAT_RGBA8, including monochrome images. This could indeed be improved, but I think it is something for another PR (I would be happy to look into that, but I prefer this PR getting resolved first).

[Ithamar]

On another note, while I recently used Image::load_bmp_from_buffer, I notice the docs actually state that is does not support 16-bit, though it does not give a reason. @Calinou are you aware of why that is?

[Calinou]

On another note, while I recently used Image::load_bmp_from_buffer, I notice the docs actually state that is does not support 16-bit, though it does not give a reason. @Calinou are you aware of why that is?

The only mention of 16-bit BMP I can see is in #30634. I don't think there's a specific reason for 16-bit BMP not being supported, so feel free to add support for it as well 🙂

Please do so in a separate pull request, so they can be reviewed and merged independently.

[SaracenOne]

Seems fine to me!

[akien-mga]

Thanks! And congrats for your first merged Godot contribution 🎉

[timothyqiu]

Cherry-picked for 3.6