(Closes #2381) alternative approach to global reduction

[arporter]

Refactors the existing support for global sums and begins adding support for global max/min.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.91%. Comparing base (97dd02d) to head (0eea98f).

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #3222   +/-   ##
=======================================
  Coverage   99.91%   99.91%           
=======================================
  Files         376      378    +2     
  Lines       53529    53525    -4     
=======================================
- Hits        53484    53480    -4     
  Misses         45       45           

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

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[arporter]

The testing was becoming increasingly artificial as I didn't have any "legitimate" way to introduce kernels with gh_min or gh_max access. Therefore I've started looking at adding the actual builtin(s) that will use these new accesses. #2674 requests support for field_min_max so I've begun adding that. However, I've then tripped over the fact that we currently enforce the requirement for a kernel/builtin to have only 1 or 0 reductions and min-max needs 2.