Improve performance by preallocating the output buffer

[wlkh]

Thanks for a great package. I've been using this successfully many times.

We have been working on optimizing some software using this package and while profiling we noticed that it cobs.Encode/Decode uses a lot of time on allocating memory. This PR improves the performance on Encode/Decode by preallocating the output buffers leading to much less memory allocation while running. This is most noticeable when encoding larger payloads.

Benchmark before:

$ go test -bench=. --benchmem
goos: linux
goarch: amd64
pkg: github.com/dim13/cobs
cpu: Intel(R) Core(TM) Ultra 9 185H
BenchmarkCobs/32/Encode-22         	17362496	        72.44 ns/op	      40 B/op	       2 allocs/op
BenchmarkCobs/32/Decode-22         	29137506	        40.76 ns/op	      40 B/op	       2 allocs/op
BenchmarkCobs/256/Encode-22        	 2756058	       401.2 ns/op	     776 B/op	       3 allocs/op
BenchmarkCobs/256/Decode-22        	18218660	        65.89 ns/op	     264 B/op	       2 allocs/op
BenchmarkCobs/1024/Encode-22       	  723136	      1687 ns/op	    3080 B/op	       5 allocs/op
BenchmarkCobs/1024/Decode-22       	 2752242	       455.7 ns/op	    3080 B/op	       5 allocs/op
BenchmarkCobs/4096/Encode-22       	  134398	      8721 ns/op	   17672 B/op	       9 allocs/op
BenchmarkCobs/4096/Decode-22       	  673791	      1764 ns/op	   12296 B/op	       8 allocs/op
PASS
ok  	github.com/dim13/cobs	10.792s

And after:

$ go test -bench=. --benchmem
goos: linux
goarch: amd64
pkg: github.com/dim13/cobs
cpu: Intel(R) Core(TM) Ultra 9 185H
BenchmarkCobs/32/Encode-22         	21400770	        53.49 ns/op	      32 B/op	       1 allocs/op
BenchmarkCobs/32/Decode-22         	162552748	         7.383 ns/op	       0 B/op	       0 allocs/op
BenchmarkCobs/256/Encode-22        	 4238409	       280.2 ns/op	     288 B/op	       1 allocs/op
BenchmarkCobs/256/Decode-22        	19978677	        54.38 ns/op	     288 B/op	       1 allocs/op
BenchmarkCobs/1024/Encode-22       	 1000000	      1211 ns/op	    1152 B/op	       1 allocs/op
BenchmarkCobs/1024/Decode-22       	 6245422	       183.0 ns/op	    1152 B/op	       1 allocs/op
BenchmarkCobs/4096/Encode-22       	  235732	      4596 ns/op	    4864 B/op	       1 allocs/op
BenchmarkCobs/4096/Decode-22       	 2016283	       614.8 ns/op	    4864 B/op	       1 allocs/op
PASS
ok  	github.com/dim13/cobs	11.339s

Short summary of this is that:

• 32 byte encode is 1.2x faster (21400770/17362496)
• 4096 byte encode is 2.9x faster (2016283/673791)
• 32 byte decode is 5.5x faster
• 4096 byte decode is 2.9x faster