Using a map made up of 32 buckets, with each bucket holding it's own lock.

INFO[0166] Lmap latency                                  
s1=24100
m1=37251
u100=2017
u10=2864
u1=109856
n1=55635

Based on benchmarks, a Get should take close to 1 microsecond.
In this case, u1 has 110K Gets, while u10 has only ~3000. So, that's good.
But we see a sudden jump to 37k Gets (16%), which took over 1 millisecond!
And 24k requests (~10%) took over a second to return. That's very erratic behaviour.
Overall, only 72% of Gets took under 10 microseconds.

Using my own concurrent map, without mutex locks is *really* *really* fast. Able to
load a million edges in under 30 seconds. Memory usage is now a concern.
