Make snapshot load faster

[lzydmxy]

Which issues of this PR fixes:

This PR try to fix #221
For nodes count 58680021, snapshot size 12GB

	Before(sec)	After(sec)
Parse from disk	90	70
Build children path	90	9

Change log:

Optimizations:

1. Reduced one buffer copy and destruction.
2. Storage edges in bucketed based on the parent node's bucket number, and built childrenSet in parallel.
3. Reserved space for childrenSet in advance.

[JackyWoo]

from_zk_snapshot is omited

[JackyWoo]

@lzydmxy Nice optimization! Please describe your solution.

[JackyWoo]

Is there any optimization for parsing data from disk? For I found it is faster in your benchmark.

[lzydmxy]

@lzydmxy Nice optimization! Please describe your solution.
Here are three optimization points:
Reduced one buffer copy and destruction.
Storage edges in bucketed based on the parent node's bucket number, and built childrenSet in parallel.
Reserved space for childrenSet in advance.

[JackyWoo]

Please throw Exception and add an error code

[lzydmxy]

@JackyWoo The test failed for apple clang, should we continue support it ?

Run export CC=`which clang` CXX=`which clang++` && cmake -G Ninja -B ./build -DCMAKE_BUILD_TYPE=Release
-- CMAKE_TOOLCHAIN_FILE - 
-- The C compiler identification is AppleClang 15.0.0.15000040
-- The CXX compiler identification is AppleClang 15.0.0.15000040

[JackyWoo]

We should not use Apple clang. Let us update the CI env.