benchmarks for write with compression enabled/disabled

[FZambia]

This is to help in investigating #203

Output on my machine:

BenchmarkWriteNoCompression-4            3000000               503 ns/op              48 B/op          1 allocs/op
BenchmarkWriteWithCompression-4            30000             41481 ns/op             123 B/op          3 allocs/op

Investingating only BenchmarkWriteWithCompression, here is cpu profile output:

(pprof) top 10 --cum
1.14s of 1.55s total (73.55%)
Showing top 10 nodes out of 67 (cum >= 0.95s)
      flat  flat%   sum%        cum   cum%
         0     0%     0%      1.54s 99.35%  github.com/FZambia/websocket.(*Conn).WriteMessage
         0     0%     0%      1.54s 99.35%  github.com/FZambia/websocket.BenchmarkWriteWithCompression
         0     0%     0%      1.54s 99.35%  runtime.goexit
         0     0%     0%      1.54s 99.35%  testing.(*B).launch
         0     0%     0%      1.54s 99.35%  testing.(*B).runN
     0.02s  1.29%  1.29%      1.21s 78.06%  github.com/FZambia/websocket.(*Conn).NextWriter
     1.08s 69.68% 70.97%      1.08s 69.68%  runtime.memclr
     0.03s  1.94% 72.90%      1.03s 66.45%  github.com/FZambia/websocket.compressNoContextTakeover
         0     0% 72.90%      0.95s 61.29%  compress/flate.(*Writer).Reset
     0.01s  0.65% 73.55%      0.95s 61.29%  compress/flate.(*compressor).reset

(pprof) list NextWriter
Total: 1.55s
ROUTINE ======================== github.com/FZambia/websocket.(*Conn).NextWriter in /Users/a.emelin/work/gopath/src/github.com/FZambia/websocket/conn.go
      20ms      1.21s (flat, cum) 78.06% of Total
         .          .    444:   }
         .          .    445:
         .          .    446:   mw := &messageWriter{
         .          .    447:           c:         c,
         .          .    448:           frameType: messageType,
      20ms      180ms    449:           pos:       maxFrameHeaderSize,
         .          .    450:   }
         .          .    451:   c.writer = mw
         .          .    452:   if c.newCompressionWriter != nil && c.enableWriteCompression && isData(messageType) {
         .      1.03s    453:           w := c.newCompressionWriter(c.writer)
         .          .    454:           mw.compress = true
         .          .    455:           c.writer = w
         .          .    456:   }
         .          .    457:   return c.writer, nil
         .          .    458:}
(pprof) list compressNoContextTakeover
Total: 1.55s
ROUTINE ======================== github.com/FZambia/websocket.compressNoContextTakeover in /Users/a.emelin/work/gopath/src/github.com/FZambia/websocket/compression.go
      30ms      1.03s (flat, cum) 66.45% of Total
         .          .     33:   fr.(flate.Resetter).Reset(io.MultiReader(r, strings.NewReader(tail)), nil)
         .          .     34:   return &flateReadWrapper{fr}
         .          .     35:}
         .          .     36:
         .          .     37:func compressNoContextTakeover(w io.WriteCloser) io.WriteCloser {
      10ms       40ms     38:   tw := &truncWriter{w: w}
         .       10ms     39:   fw, _ := flateWriterPool.Get().(*flate.Writer)
         .      950ms     40:   fw.Reset(tw)
      20ms       30ms     41:   return &flateWriteWrapper{fw: fw, tw: tw}
         .          .     42:}
         .          .     43:
         .          .     44:// truncWriter is an io.Writer that writes all but the last four bytes of the
         .          .     45:// stream to another io.Writer.
         .          .     46:type truncWriter struct {

Heap:

$ go tool pprof --alloc_space websocket.test mem
Entering interactive mode (type "help" for commands)
(pprof) top 10 --cum
6.53MB of 8.64MB total (75.52%)
Showing top 10 nodes out of 23 (cum >= 5.13MB)
      flat  flat%   sum%        cum   cum%
         0     0%     0%     8.64MB   100%  runtime.goexit
    1.50MB 17.36% 17.36%     8.13MB 94.03%  github.com/FZambia/websocket.(*Conn).NextWriter
         0     0% 17.36%     8.13MB 94.03%  github.com/FZambia/websocket.(*Conn).WriteMessage
         0     0% 17.36%     8.13MB 94.03%  github.com/FZambia/websocket.BenchmarkWriteWithCompression
         0     0% 17.36%     8.13MB 94.03%  testing.(*B).runN
         0     0% 17.36%     7.24MB 83.83%  testing.(*B).launch
    1.50MB 17.36% 34.72%     6.63MB 76.67%  github.com/FZambia/websocket.compressNoContextTakeover
    3.53MB 40.80% 75.52%     5.13MB 59.31%  compress/flate.NewWriter
         0     0% 75.52%     5.13MB 59.31%  github.com/FZambia/websocket.glob..func1
         0     0% 75.52%     5.13MB 59.31%  sync.(*Pool).Get

(pprof) list NextWriter
Total: 8.64MB
ROUTINE ======================== github.com/FZambia/websocket.(*Conn).NextWriter in /Users/a.emelin/work/gopath/src/github.com/FZambia/websocket/conn.go
    1.50MB     8.13MB (flat, cum) 94.03% of Total
         .          .    444:   }
         .          .    445:
         .          .    446:   mw := &messageWriter{
         .          .    447:           c:         c,
         .          .    448:           frameType: messageType,
    1.50MB     1.50MB    449:           pos:       maxFrameHeaderSize,
         .          .    450:   }
         .          .    451:   c.writer = mw
         .          .    452:   if c.newCompressionWriter != nil && c.enableWriteCompression && isData(messageType) {
         .     6.63MB    453:           w := c.newCompressionWriter(c.writer)
         .          .    454:           mw.compress = true
         .          .    455:           c.writer = w
         .          .    456:   }
         .          .    457:   return c.writer, nil
         .          .    458:}
(pprof) list compressNoContextTakeover
Total: 8.64MB
ROUTINE ======================== github.com/FZambia/websocket.compressNoContextTakeover in /Users/a.emelin/work/gopath/src/github.com/FZambia/websocket/compression.go
    1.50MB     6.63MB (flat, cum) 76.67% of Total
         .          .     33:   fr.(flate.Resetter).Reset(io.MultiReader(r, strings.NewReader(tail)), nil)
         .          .     34:   return &flateReadWrapper{fr}
         .          .     35:}
         .          .     36:
         .          .     37:func compressNoContextTakeover(w io.WriteCloser) io.WriteCloser {
  512.02kB   512.02kB     38:   tw := &truncWriter{w: w}
         .     5.13MB     39:   fw, _ := flateWriterPool.Get().(*flate.Writer)
         .          .     40:   fw.Reset(tw)
       1MB        1MB     41:   return &flateWriteWrapper{fw: fw, tw: tw}
         .          .     42:}
         .          .     43:
         .          .     44:// truncWriter is an io.Writer that writes all but the last four bytes of the
         .          .     45:// stream to another io.Writer.
         .          .     46:type truncWriter struct {