[Feat][Doc] Benchmarking C++/Spark readers with/without filter pushdown

[lixueclaire]

Is your feature request related to a problem? Please describe.
Filter pushdown is a sophisticated feature available with C++/Spark readers that has the potential to enhance query performance. It's important to evaluate its effectiveness.

Describe the solution you'd like
To gauge the impact of filter pushdown, I propose using the LDBC dataset to benchmark the performance of reading operations. Specifically, we can measure how efficiently the C++/Spark readers can filter vertices or edges with certain property conditions when filter pushdown is enabled compared to when it is not.

Additional context
This request is in continuation of the discussion in issue #389

[SemyonSinchenko]

I want to start working on Spark benchmark. What do you think guys, should it be a separate project or not? I want to start from basic read/write, read with pushdown, ego-nets.

I see it as a JMH-based benchmark cases, that uses the same data like we are using for tests.