In place XOR

[DhruvilK7]

It is almost at same performance if the second bitmap is much sparser than the first or two random bitmaps with the same cardinality and range. However it is much faster for two dense bitmaps.

Fulfills part of #27.

Benchmark:

benchmark                           old ns/op     new ns/op     delta
BenchmarkXor-11                     785310        775829        -1.21%
BenchmarkXorLopsided-11             89268         89067         -0.23%
BenchmarkXorDense-11                116055        64905         -44.07%

benchmark                           old allocs    new allocs    delta
BenchmarkXor-11                     6108          6108          +0.00%
BenchmarkXorLopsided-11             3120          3120          +0.00%
BenchmarkXorDense-11                100           0             -100.00%

benchmark                           old bytes     new bytes     delta
BenchmarkXor-11                     513732        513733        +0.00%
BenchmarkXorLopsided-11             210225        210225        +0.00%
BenchmarkXorDense-11                411204        0             -100.00%

[lemire]

Please run go fmt if you can.

[DhruvilK7]

done