refactor(kernels::grisubal): rewrite step 1 to enable parallelization

[imrn99]

there are two main changes to make it happen:

• split the main iterator in two, to compute and preallocate intersection metadata vector, allowing edition of the vector with no pushes
• rewrite the second part of the iterator to collect (K, V) pairs directly into the Hashmap rather than insert each manually

using the bench scripts, it seems like the performance is unchanged for now, which is a win since step 1 is now parallelizable

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 82.70%. Comparing base (d79671f) to head (549bd42).
Report is 2 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #210      +/-   ##
==========================================
+ Coverage   82.04%   82.70%   +0.66%     
==========================================
  Files          41       41              
  Lines        5681     5898     +217     
==========================================
+ Hits         4661     4878     +217     
  Misses       1020     1020              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[imrn99]

I think I will split the kernel.rs file into a module in a separate PR. the file is starting to be very long, and each step are starting to contain many substeps