Relevant updates:

* Less allocations on decode: https://github.com/klauspost/compress/pull/322
* Fixed rare out-of-bounds write on amd64.
* ARM64 decompression assembly. Around 2x output speed. https://github.com/klauspost/compress/pull/324
* Speed up decompression on non-assembly platforms. https://github.com/klauspost/compress/pull/328

Upgrade cpuid to match simdjson.