quicvarint: add a function to parse a varint from a byte slice

[marten-seemann]

Part of #4477.

The performance difference is quite large, especially for longer varints:

goos: darwin
goarch: arm64
pkg: github.com/quic-go/quic-go/quicvarint
BenchmarkRead
BenchmarkRead/1-byte
BenchmarkRead/1-byte-16         	361651708	         3.380 ns/op
BenchmarkRead/2-byte
BenchmarkRead/2-byte-16         	257981850	         4.648 ns/op
BenchmarkRead/4-byte
BenchmarkRead/4-byte-16         	168197124	         7.102 ns/op
BenchmarkRead/8-byte
BenchmarkRead/8-byte-16         	96493723	        12.29 ns/op
BenchmarkParse
BenchmarkParse/1-byte
BenchmarkParse/1-byte-16        	514354692	         2.195 ns/op
BenchmarkParse/2-byte
BenchmarkParse/2-byte-16        	522608250	         2.244 ns/op
BenchmarkParse/4-byte
BenchmarkParse/4-byte-16        	556674920	         2.165 ns/op
BenchmarkParse/8-byte
BenchmarkParse/8-byte-16        	497475228	         2.449 ns/op

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 84.94%. Comparing base (0d565fc) to head (3fa2807).
Report is 1 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #4475      +/-   ##
==========================================
+ Coverage   84.88%   84.94%   +0.07%     
==========================================
  Files         154      154              
  Lines       14647    14685      +38     
==========================================
+ Hits        12432    12474      +42     
+ Misses       1702     1697       -5     
- Partials      513      514       +1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.