Thank you for taking a stab at this!
We actually don't need this new fill_buffer method in wgpu-hal, there is already clear_buffer for exactly the same purpose. The API difference between them is the value: u8, but if you look at WebGPU spec, there isn't a value there either (because it would be more difficult to implement) - https://gpuweb.github.io/gpuweb/

So all that needs to be done here is exposing the method to wgpu-rs

Ah, alright, I misread part of the discussions over gpuweb haha.
So this would be more of a renaming from clear_buffer to fill_buffer, wouldn't it?

We had it called fill_buffer at some point, then renamed to clear_buffer. I guess it's time to switch it back now :)
cc @Wumpf who did most of this work.

Another thing is - this API should now be available without any features, previously it was behind a feature.

oh cool didn't know that this is now in the spec. Nice!
Wanted to change the meaning of the extension a little bit towards "this is what zero init does when it hits", but that would change that again then (:
Wumpf@127cbab

Alright, but does clear_texture remain under CLEAR_COMMANDS?

Also, by gpuweb/gpuweb#2208 comment, value may be added in the future (is on the workings I think: gpuweb/gpuweb#2290) so, do you think is better to stall this PR until then and create another one to keep up with the spec in the meantime?

Alright, but does clear_texture remain under CLEAR_COMMANDS?

yes, and we should rename it to CLEAR_TEXTURES, but I can do that later; going to make some changes to it anyways.
I'd say let's add value when it enters the spec. With zero we can implement it on all backends today, but adding a value means that we need to solve this with shaders on some backends

Okay, so I removed all the previous commits from this branch. They will still remain on my branch (https://github.com/UpsettingBoy/wgpu/tree/cmd-encoder-value) so when the spec adds value there'll be some work done :)

@Wumpf I left CLEAR_COMMAND untouched for you to change.