more i16 and u16 operations

[lovasoa]

Hello,
I was hoping I could use this crate for image-rs/jpeg-decoder#146 , but it would require several functions that are not implemented:

• cast from i16 to i32
• cast from u16 to i32
• load i16
• load u16

[jackmott]

On the unsigned 16s, do you need all the add/sub etc functions for those too? There isn't any u16 stuff at the moment.

[jackmott]

I've added the load i16, the cast I need to do some research on how to do it in sse2 efficiently. PRs welcome by the way if you want to add these yourself. I can help you get started if you don't know where to begin.

[lovasoa]

Thanks for implementing load_i16 !
The code I'm working on (JPEG IDCT) just loads i16 and u16, then casts them to i32, then performs arithmetic, so it doesn't need any operation on i16 and u16...

Yes, I realized there aren't any u16 function at the moment after I created the issue. It's too bad because the JPEG quantization table is made of u16s...

I may try to open a PR if I find the time (and the motivation 😬 ).