Optimize node encoding/decoding with stream-vbyte format

[yihuang]

This is a node storage format change, need to migrate database, but I guess it's ok in light of new node key format, we'll need to migrate anyway, the same technique can be applied to new node key format as well.
The improvement is pretty good, we don't even use the SIMP optimizations, since we don't have that much integers.

Downside:

• don't support version bigger than math.MaxUint32.
  • I did a calculation, say block time is 2 second, (((1<<32)-1) * 2) / 3600/24/365 = 272, it means it takes 272 years for the version number to overflow uint32, we are using 6 seconds block time right now, which makes it even longer.

Benchmark

$ go test -run=^$ -bench="maxEncodedSize|Encode|Decode" -benchmem ./ -count=10

Compare with this version where I adjusted the test cases the same as this PR:

name                          old time/op    new time/op    delta
Node_maxEncodedSize-12          7.33ns ± 4%    0.24ns ± 3%  -96.67%  (p=0.000 n=10+9)
Node_Encode/small_integer-12     245ns ± 1%      55ns ± 1%  -77.62%  (p=0.000 n=10+10)
Node_Encode/large_integer-12     245ns ± 2%      55ns ± 0%  -77.64%  (p=0.000 n=10+9)
Node_Decode/small_number-12      169ns ± 1%      65ns ± 2%  -61.73%  (p=0.000 n=10+9)
Node_Decode/large_number-12      170ns ± 1%      65ns ± 1%  -61.92%  (p=0.000 n=10+9)

name                          old alloc/op   new alloc/op   delta
Node_maxEncodedSize-12           0.00B          0.00B          ~     (all equal)
Node_Encode/small_integer-12      160B ± 0%      112B ± 0%  -30.00%  (p=0.000 n=10+10)
Node_Encode/large_integer-12      160B ± 0%      112B ± 0%  -30.00%  (p=0.000 n=10+10)
Node_Decode/small_number-12       256B ± 0%      160B ± 0%  -37.50%  (p=0.000 n=10+10)
Node_Decode/large_number-12       256B ± 0%      160B ± 0%  -37.50%  (p=0.000 n=10+10)

name                          old allocs/op  new allocs/op  delta
Node_maxEncodedSize-12            0.00           0.00          ~     (all equal)
Node_Encode/small_integer-12      2.00 ± 0%      1.00 ± 0%  -50.00%  (p=0.000 n=10+10)
Node_Encode/large_integer-12      2.00 ± 0%      1.00 ± 0%  -50.00%  (p=0.000 n=10+10)
Node_Decode/small_number-12       4.00 ± 0%      1.00 ± 0%  -75.00%  (p=0.000 n=10+10)
Node_Decode/large_number-12       4.00 ± 0%      1.00 ± 0%  -75.00%  (p=0.000 n=10+10)

[yihuang]

I'll close it for now, no resource to pursue it for now.

[tac0turtle]

Would love to open an issue for this. It looks awesome, maybe someone from outside can pick it up