Introduce an interface RWLock to implement lock free in production mode

[lunny]

Improvement after #90 #92

goos: darwin
goarch: amd64
pkg: github.com/unrolled/render
cpu: Intel(R) Core(TM) i7-1068NG7 CPU @ 2.30GHz
(Before)BenchmarkHTML-8          1341072               892.2 ns/op
(After)   BenchmarkHTML-8          1374212               855.5 ns/op
PASS

[zeripath]

I think Rlocks are really very fast to take out especially if there isn't ever a write lock.

It would be good to benchmark this to prove that it does make a difference.

[zeripath]

So running the benchmarks it looks like surprisingly this is slower. But I think it's all within the bounds of error.

master	BenchmarkBigHTMLBuffers-12    	  933356	      1240 ns/op	     376 B/op	      11 allocs/op
91	BenchmarkBigHTMLBuffers-12    	  912496	      1220 ns/op	     376 B/op	      11 allocs/op
92 	BenchmarkBigHTMLBuffers-12    	  879910	      1223 ns/op	     376 B/op	      11 allocs/op
93 	BenchmarkBigHTMLBuffers-12    	  946242	      1255 ns/op	     376 B/op	      11 allocs/op
94 	BenchmarkBigHTMLBuffers-12    	  915399	      1258 ns/op	     376 B/op	      11 allocs/op

master	BenchmarkSmallHTMLBuffers-12    	  862422	      1407 ns/op	     464 B/op	      14 allocs/op
91	BenchmarkSmallHTMLBuffers-12    	  866845	      1327 ns/op	     464 B/op	      14 allocs/op
92	BenchmarkSmallHTMLBuffers-12    	  864735	      1316 ns/op	     464 B/op	      14 allocs/op
93	BenchmarkSmallHTMLBuffers-12    	  835960	      1316 ns/op	     464 B/op	      14 allocs/op
94	BenchmarkSmallHTMLBuffers-12    	  838126	      1352 ns/op	     464 B/op	      14 allocs/op

[lunny]

When I test with 5s, another suprsing result.

go test -bench=BenchmarkHTML -cpu=12 -benchtime=5s

goos: darwin
goarch: amd64
pkg: github.com/unrolled/render
cpu: Intel(R) Core(TM) i7-1068NG7 CPU @ 2.30GHz
(Before) BenchmarkHTML-12         7135489               817.8 ns/op
(After)   BenchmarkHTML-12         7400484               776.3 ns/op
PASS
ok      github.com/unrolled/render      6.716s

[zeripath]

I get:

master  BenchmarkHTML-12    	 9416479	       654.9 ns/op
#91	BenchmarkHTML-12    	 9254259	       676.1 ns/op
#92	BenchmarkHTML-12    	 9732022	       645.7 ns/op
#93	BenchmarkHTML-12    	 9386546	       651.7 ns/op
#94	BenchmarkHTML-12    	 9203570	       656.3 ns/op

The small differences are probably because the templates contains a lock itself - which we can't really escape from.

[unrolled]

Can you merge master into this PR?

[lunny]

Can you merge master into this PR?

Did you mean branch v1?

[unrolled]

Thanks for the PR!