OpenZL: Introduce a Go based implementation

[marclop]

Description

Meta just released a new novel compression algorithm called OpenZL (see blog post).

Many applications would likely benefit from using OpenZL since the schema of the data being compressed is known in advance. OpenZL has been shown to improve the compression ratio even from Zstandard (zstd) in these cases.

I'm opening this issue to see if you're interested in implementing a pure Go-based implementation of OpenZL, although it's just launched, I can see how many may want to adopt this.

Big fan of your work and this repo. We (at Elastic) are using this repo and the implementations here in a few places.

[klauspost]

I don't know how deeply you looked into it - but (for others reading) it is not a generic compressor. It is more like a toolbox where you can write custom compressors for data with fixed schemas. The main advantage is that you don't also have to write all the plumbing nor a decompressor.

So it is domain-specific compression and in particular the "no recursion restrict" - and for it to be reasonably effective you need to write the custom splitters. The current filters are enough to cover some reasonable starting cases, but it seems they are very much expecting to break existing and add more for the v1 release.

As such, I am not sure it is worth starting to port at such an early stage. I feel that CGO bindings would be more reasonable for testing out, while the format is stabilizing. I do not currently plan to do bindings.

When things are looking stable, I will probably add it as separate package.