Benchmark regression alp

[seb711]

This is a super bulky PR; therefore it is only in draft. I think it needs a few adjustments until it can be merged. Following a list of thoughts and things that need to be done:

• for testing and benchmarking reasons I changed decompression-speed and csvtobtr => new files should be created with their content
• ALP and PFOR (for int64) was added and checked with the PBI dataset; i also included a file to benchmark/test the implementation (tools/regression-benchmark/column_benchmark.cpp) => this tests compression/decompression and checks the contents after decompression; it also outputs PerfEvent measurement
• the INT64 schemes and architecture is the work of Maximilian Kuschewski and it is not sure if there could be more work done to template the schemes/refactorment
• for completeness I included also the AlpRD implementation; in my tests AlpRD was throwing an SEGFAULT with the dataset Bimbo/1/Bimbo_1/5_Dev_proxima.double; therefore i disabled it for now (i didnt had the time to look into it)
• also the implementation of FastPFOR int64 is very sparse and opted for working

otherwise this PR works and you can now use integer64 values and compress your data with the novel ALP approach; tests yield that you get way better compression factors for double values.

short note on the ALP implementation:

• initially i compressed the whole block of 2^16 values. after consideration of performance i modified the logic so that the compression happens batchwise with batches of size 2^12. i also tested a bit with simd (see commit history), but on the one hand i couldnt really see a performance improvement in decompression and on the other side i introduced bugs when casting vectors of ints to doubles. therefore no simd at the time.