perf: reduce temp allocation in serialization

[proost]

During Serialization, temporary allocations happens due to collecting active values and active keys.

But we can remove those allocations and reduce 4 loops to one.

main branch:

goos: darwin
goarch: arm64
pkg: github.com/apache/datasketches-go/frequencies
cpu: Apple M4 Pro
BenchmarkLongsSketchToSlice
BenchmarkLongsSketchToSlice/active=48
BenchmarkLongsSketchToSlice/active=48-12         	 4521978	       262.5 ns/op	    1664 B/op	       3 allocs/op
BenchmarkLongsSketchToSlice/active=768
BenchmarkLongsSketchToSlice/active=768-12        	  344616	      3540 ns/op	   25856 B/op	       3 allocs/op
BenchmarkLongsSketchToSlice/active=12288
BenchmarkLongsSketchToSlice/active=12288-12      	   20967	     56307 ns/op	  401410 B/op	       3 allocs/op
BenchmarkLongsSketchToSlice/active=196608
BenchmarkLongsSketchToSlice/active=196608-12     	    1006	   1112128 ns/op	 6299650 B/op	       3 allocs/op
PASS

This PR:

goos: darwin
goarch: arm64
pkg: github.com/apache/datasketches-go/frequencies
cpu: Apple M4 Pro
BenchmarkLongsSketchToSlice
BenchmarkLongsSketchToSlice/active=48
BenchmarkLongsSketchToSlice/active=48-12         	 9544124	       126.4 ns/op	     896 B/op	       1 allocs/op
BenchmarkLongsSketchToSlice/active=768
BenchmarkLongsSketchToSlice/active=768-12        	  701419	      1667 ns/op	   13568 B/op	       1 allocs/op
BenchmarkLongsSketchToSlice/active=12288
BenchmarkLongsSketchToSlice/active=12288-12      	   38566	     31243 ns/op	  204803 B/op	       1 allocs/op
BenchmarkLongsSketchToSlice/active=196608
BenchmarkLongsSketchToSlice/active=196608-12     	    2101	    559105 ns/op	 3153921 B/op	       1 allocs/op
PASS