[sz3]

Definitely. I had to draw a line somewhere for an initial "does anyone care about this?" release, so there's some rough edges here.

There are a number of loose ends that I'd like to look into down the road, and more flexibility for different use cases is on the list.

edit: I should mention: the main reason the decoder is a native Android app and not a WebAssembly app is that decoding performance is a throughput bottleneck. I wasn't too eager to pay the wasm tax, when I wasn't sure if even native performance would be good enough. As it happens, I now think decode performance is going to be ok -- but it's still a bit of a weak spot in the scheme.