[bugfix] Replace custom direction filter for loopback interfaces with BPF filter

[fako1024]

@els0r I'm glad this popped up, gave me the chance to rethink the strategy and drop the whole internal filtering complexity in favor of the far more efficient BPF filter logic on kernel level. See Issue comment in #75 for more details:

Benchmarks:

                                                 │ nobpf │            bpf            │
                                                 │     sec/op     │    sec/op     vs base               │
CaptureMethods/NextPacket_10485kiBx4-4               78.64n ±  0%   77.22n ±  0%  -1.81% (p=0.000 n=15)
CaptureMethods/NextPacketInPlace_10485kiBx4-4        34.99n ±  1%   34.78n ±  0%  -0.60% (p=0.002 n=15)
CaptureMethods/NextPayload_10485kiBx4-4              69.24n ±  0%   70.26n ±  0%  +1.47% (p=0.000 n=15)
CaptureMethods/NextPayloadInPlace_10485kiBx4-4       29.83n ±  1%   29.66n ±  1%       ~ (p=0.237 n=15)
CaptureMethods/NextPayloadZeroCopy_10485kiBx4-4      17.09n ±  1%   17.05n ±  1%       ~ (p=0.418 n=15)
CaptureMethods/NextIPPacket_10485kiBx4-4             73.32n ±  0%   71.40n ±  0%  -2.62% (p=0.000 n=15)
CaptureMethods/NextIPPacketInPlace_10485kiBx4-4      34.51n ±  1%   32.76n ±  1%  -5.07% (p=0.000 n=15)
CaptureMethods/NextIPPacketZeroCopy_10485kiBx4-4     17.21n ±  1%   17.04n ±  1%  -0.99% (p=0.004 n=15)
CaptureMethods/NextPacketFn_10485kiBx4-4             24.25n ± 11%   23.33n ± 10%       ~ (p=0.267 n=15)
CaptureMethods/NextPacket_10kiBx512-4                205.0n ±  7%   196.5n ±  1%  -4.15% (p=0.000 n=15)
CaptureMethods/NextPacketInPlace_10kiBx512-4         143.4n ±  3%   139.4n ±  1%  -2.79% (p=0.008 n=15)
CaptureMethods/NextPayload_10kiBx512-4               183.4n ±  1%   177.0n ±  2%  -3.49% (p=0.000 n=15)
CaptureMethods/NextPayloadInPlace_10kiBx512-4        144.2n ±  3%   143.3n ±  4%       ~ (p=0.767 n=15)
CaptureMethods/NextPayloadZeroCopy_10kiBx512-4       149.9n ±  2%   143.7n ±  1%  -4.14% (p=0.000 n=15)
CaptureMethods/NextIPPacket_10kiBx512-4              183.9n ±  3%   178.1n ±  2%       ~ (p=0.123 n=15)
CaptureMethods/NextIPPacketInPlace_10kiBx512-4       141.6n ±  2%   139.3n ±  2%  -1.62% (p=0.012 n=15)
CaptureMethods/NextIPPacketZeroCopy_10kiBx512-4      149.6n ±  3%   142.6n ±  2%  -4.68% (p=0.000 n=15)
CaptureMethods/NextPacketFn_10kiBx512-4              145.8n ±  2%   142.2n ±  3%  -2.47% (p=0.012 n=15)
geomean                                              75.78n         74.03n        -2.31%

Closes #75

[els0r]

See minor comment about switch statement.