0.7.1

This release include another full rewrite of the tag parser, using a new approach that resulted
in an average 50% performance improvement over version 0.7.0.

# Baseline: f5c6c32da475a7436c0aa58e4f24874364955dcf

# ARM NEON
twitch/1000 
  before: 245.584 µs
  after:  121.391 µs
  change: -49.4%

# x86 AVX512
twitch/1000
  before: 188.064 µs
  after:   94.260 µs
  change: -50.1%

x86 now has implementations using SSE2, AVX2, and AVX512, choosing the best available at compile time.
For that reason, the crate should ideally be compiled with RUSTFLAGS="-C target-cpu=native".

Full commit range: 0.7.0..3b19a23