26 thp v2 arrow POC

[chrisbc]

FYI @chrisdicaprio this branch has my workings on doing more data-wrangling using arrow/numpy/pandas. Notes:

• I need your help to understand the composite/component branch data structures
• I "think" I understand how to do all the math ops in numba, but leaving implementation until above is sorted
• I expect this approach to have at least one (maybe two) orders better performance than the parent branch
• multiple IMTs will be processed together, and same approach is doable for sites. But may need to manage memory
• I've included hazard_v2_mini.toml and ..micro.toml configs for easier/faster comparisons between this (ARROW) and parent branch (OG) methods.
• new command line option --method (-M) which defaults to OG, or use ARROW to try the newer approach.
• my exploratory tests are in tests/version2/test_arrow_aggregation.py

[chrisdicaprio]

for componet_branch in composite_branch.compoent_branches:
    weight_composite *= weight