v0.2.3: more performance improvements

Another 60% improvement in decoding performance, with an 8KB memory penalty. If this is prohibitive, disable the feature:

[dependencies]
blurhash = { version = "0.2.3", default-features = false }

v0.2.2: performance improvements

This release improves decoding performance by about 60% (on a Ryzen 7840U) and encoding performance by about 70% (on the same Ryzen), by using some lookup tables for the DCT values.

If these tables are prohibitive in memory consumption, feel free to reach out and make an issue about this!