Closes #220 and #230

Closes #166

This doesn't affect real world applications due to buffering but
the testss would occasionally fail on CI due to the code not handling
an immediate close after writing the close frame while resetting
the write timeout.

Closes #179

On average, we are 1.75x faster than both now 

goos: linux
goarch: amd64
pkg: nhooyr.io/websocket
Benchmark_mask/2/basic-8         	263923018	         4.55 ns/op	 439.76 MB/s
Benchmark_mask/2/nhooyr-8        	200203578	         6.00 ns/op	 333.32 MB/s
Benchmark_mask/2/gorilla-8       	241622557	         4.95 ns/op	 404.38 MB/s
Benchmark_mask/2/gobwas-8        	200392592	         6.05 ns/op	 330.69 MB/s
Benchmark_mask/3/basic-8         	207684956	         5.78 ns/op	 519.07 MB/s
Benchmark_mask/3/nhooyr-8        	169297215	         7.10 ns/op	 422.49 MB/s
Benchmark_mask/3/gorilla-8       	205775799	         5.81 ns/op	 516.23 MB/s
Benchmark_mask/3/gobwas-8        	165921662	         7.23 ns/op	 415.06 MB/s
Benchmark_mask/4/basic-8         	167034886	         7.18 ns/op	 557.30 MB/s
Benchmark_mask/4/nhooyr-8        	287656454	         4.30 ns/op	 931.11 MB/s
Benchmark_mask/4/gorilla-8       	166140434	         7.30 ns/op	 547.57 MB/s
Benchmark_mask/4/gobwas-8        	138138087	         8.73 ns/op	 458.20 MB/s
Benchmark_mask/8/basic-8         	121179904	         9.92 ns/op	 806.67 MB/s
Benchmark_mask/8/nhooyr-8        	199632992	         6.07 ns/op	1318.60 MB/s
Benchmark_mask/8/gorilla-8       	100000000	        10.8 ns/op	 739.65 MB/s
Benchmark_mask/8/gobwas-8        	157898031	         7.54 ns/op	1061.27 MB/s
Benchmark_mask/16/basic-8        	73648268	        16.5 ns/op	 971.56 MB/s
Benchmark_mask/16/nhooyr-8       	186871615	         6.38 ns/op	2506.61 MB/s
Benchmark_mask/16/gorilla-8      	72374540	        16.6 ns/op	 964.36 MB/s
Benchmark_mask/16/gobwas-8       	127698723	         9.36 ns/op	1709.99 MB/s
Benchmark_mask/32/basic-8        	40010325	        29.8 ns/op	1073.76 MB/s
Benchmark_mask/32/nhooyr-8       	168590156	         7.12 ns/op	4494.72 MB/s
Benchmark_mask/32/gorilla-8      	67282072	        17.7 ns/op	1808.59 MB/s
Benchmark_mask/32/gobwas-8       	120038877	         9.96 ns/op	3213.55 MB/s
Benchmark_mask/128/basic-8       	10134963	       118 ns/op	1082.74 MB/s
Benchmark_mask/128/nhooyr-8      	100000000	        11.8 ns/op	10852.23 MB/s
Benchmark_mask/128/gorilla-8     	45452385	        26.4 ns/op	4853.64 MB/s
Benchmark_mask/128/gobwas-8      	57188290	        20.8 ns/op	6153.80 MB/s
Benchmark_mask/512/basic-8       	 2707371	       442 ns/op	1159.38 MB/s
Benchmark_mask/512/nhooyr-8      	37049421	        32.4 ns/op	15785.82 MB/s
Benchmark_mask/512/gorilla-8     	19006171	        62.8 ns/op	8150.01 MB/s
Benchmark_mask/512/gobwas-8      	21394864	        55.8 ns/op	9169.49 MB/s
Benchmark_mask/4096/basic-8      	  346566	      3467 ns/op	1181.46 MB/s
Benchmark_mask/4096/nhooyr-8     	 5170425	       232 ns/op	17648.07 MB/s
Benchmark_mask/4096/gorilla-8    	 2963664	       405 ns/op	10105.59 MB/s
Benchmark_mask/4096/gobwas-8     	 2989596	       402 ns/op	10192.40 MB/s
Benchmark_mask/16384/basic-8     	   86671	     13833 ns/op	1184.38 MB/s
Benchmark_mask/16384/nhooyr-8    	 1332649	       889 ns/op	18436.40 MB/s
Benchmark_mask/16384/gorilla-8   	  763900	      1556 ns/op	10527.53 MB/s
Benchmark_mask/16384/gobwas-8    	  764034	      1553 ns/op	10548.60 MB/s
PASS
ok  	nhooyr.io/websocket	64.091s

Results from a 8 GB 8 core Haswell VM on GCP.

Much faster.

Subprotocols didn't work with WASM at all before.

Closes #121

The boys in the office hated it.

Highly unfortunate but makes sense as it seems confusing.

Will be cleaner in GitHub as the README.md is the first
thing you will see instead of a massive directory listing.

Closes #134

On top of the new Go tests designed to
maximize coverage.

Just so I can get rid of the autobahn-testsuite with confidence.

Many new helpers.

Closes #124

Race detector doesn't work on alpine.

See https://github.com/golang/go/issues/14481

Closes #26