Use of SIMD computation in CRHMC walk

[iakoviid]

trafficstars

Expanding on PR #239 we use the SIMD capabilities of PackedCSparse and present a vectorized CRHMC walk. This way we can have speed up on medium to large datasets, depending on the systems architecture. Changes from #239 :

1. Changes in CRHMC walk instead of using n-dimensional vectors we use n * simd_len-dimensional matrices. In this way we do simd_len parallel CRHMC walks without spawning additional threads.
2. Changes in testing and benchmarking files to test the behavior of the added code.

[codecov[bot]]

Codecov Report

Merging #245 (03a08c5) into develop (074a562) will decrease coverage by 0.15%. The diff coverage is 47.48%.

:exclamation: Current head 03a08c5 differs from pull request most recent head 1c03920. Consider uploading reports for the commit 1c03920 to get more accurate results

Additional details and impacted files

@@             Coverage Diff             @@
##           develop     #245      +/-   ##
===========================================
- Coverage    54.67%   54.52%   -0.15%     
===========================================
  Files          107      106       -1     
  Lines         6337     6369      +32     
  Branches      2942     2958      +16     
===========================================
+ Hits          3465     3473       +8     
+ Misses         962      958       -4     
- Partials      1910     1938      +28     

Files	Coverage Δ
include/preprocess/crhmc/opts.h	0.00% <ø> (ø)
include/ode_solvers/oracle_functors.hpp	28.57% <50.00%> (ø)
include/preprocess/crhmc/crhmc_input.h	80.43% <86.66%> (+4.67%)	:arrow_up:
include/preprocess/crhmc/analytic_center.h	28.20% <25.00%> (+4.82%)	:arrow_up:
...random_walks/crhmc/additional_units/auto_tuner.hpp	53.33% <25.00%> (+6.66%)	:arrow_up:
include/preprocess/crhmc/lewis_center.h	20.93% <0.00%> (+0.93%)	:arrow_up:
include/preprocess/crhmc/two_sided_barrier.h	60.93% <84.00%> (+24.93%)	:arrow_up:
include/sampling/random_point_generators.hpp	72.22% <75.00%> (+2.22%)	:arrow_up:
include/ode_solvers/implicit_midpoint.hpp	58.33% <37.50%> (+0.70%)	:arrow_up:
...lude/preprocess/crhmc/weighted_two_sided_barrier.h	60.00% <71.42%> (+5.00%)	:arrow_up:
... and 9 more

... and 9 files with indirect coverage changes

[vissarion]

@iakoviid there are some issues still in CI tests (most probably new). There is one SEGFAULT with clang-12 and an error in R tests in windows (where some psrf is NaN). Could you please check?

[TolisChal]

I'm closing this PR since it is included in https://github.com/GeomScale/volesti/pull/286