mat64: add r/w benchmarks for Dense and Vector

[sbinet]

BenchmarkMarshalDense10-4         500000          2635 ns/op         400 B/op         26 allocs/op
BenchmarkMarshalDense100-4        100000         22186 ns/op        2640 B/op        206 allocs/op
BenchmarkMarshalDense1000-4        10000        218528 ns/op       24336 B/op       2006 allocs/op
BenchmarkMarshalDense10000-4        1000       2173997 ns/op      242068 B/op      20006 allocs/op
BenchmarkUnmarshalDense10-4      2000000           914 ns/op         272 B/op          8 allocs/op
BenchmarkUnmarshalDense100-4      500000          3808 ns/op        1904 B/op          8 allocs/op
BenchmarkUnmarshalDense1000-4      50000         32730 ns/op       16496 B/op          8 allocs/op
BenchmarkUnmarshalDense10000-4      5000        310645 ns/op      163955 B/op          8 allocs/op
BenchmarkMarshalVector10-4       1000000          2467 ns/op         384 B/op         24 allocs/op
BenchmarkMarshalVector100-4       100000         21861 ns/op        2624 B/op        204 allocs/op
BenchmarkMarshalVector1000-4       10000        214783 ns/op       24320 B/op       2004 allocs/op
BenchmarkMarshalVector10000-4       1000       2155109 ns/op      242052 B/op      20004 allocs/op
BenchmarkUnmarshalVector10-4     2000000           813 ns/op         256 B/op          6 allocs/op
BenchmarkUnmarshalVector100-4     500000          3701 ns/op        1888 B/op          6 allocs/op
BenchmarkUnmarshalVector1000-4     50000         32248 ns/op       16480 B/op          6 allocs/op
BenchmarkUnmarshalVector10000-4     5000        311020 ns/op      163939 B/op          6 allocs/op

Updates #346

[jonlawlor]

By the way I am almost done with the first pass at benchmark plotting. Ended up using d3 javascript instead of gonum plot so you can get pretty interactive plots. So thanks for using a number for the size in these benchmarks.

[kortschak]

LGTM