related_post_gen benchmark

[PMunch]

Okay, gave FxHash a go. Implemented it like this:
    
    
    import bitops
    from hashes import Hash
    
    const
      rotate = 5
      seed = 0x517cc1b727220a95'u64
    
    proc hashWord(x: var Hash, word: uint64) =
      x = cast[Hash]((x.uint64.rotateLeftBits(rotate) xor word) * seed)
    
    proc hash(x: string): Hash =
      var pos = 0
      while x.high - pos > 8:
        result.hashWord(cast[ptr uint64](x[pos].addr)[])
        pos += 8
      for pos in pos..x.high:
        result.hashWord(x[pos].uint64)
    
    import std/monotimes
    import hashes
    
    let s1 = getMonoTime()
    var h1: Hash
    for i in 0..100:
      h1 = h1 !& fxhash.hash("Hello world")
    echo getMonoTime() - s1
    echo h1
    
    let s2 = getMonoTime()
    var h2: Hash
    for i in 0..100:
      h2 = h2 !& hashes.hash("Hello world")
    echo getMonoTime() - s2
    echo h2
    
    
    Run

Runs the FxHash block in 267 nanoseconds and the built-in hash in 1058 
nanoseconds.

Tried it in the benchmark and it dropped the time from ~75ms to ~53ms, so 
definitely a speedup!