Process rep nodes when collapsing an SCC

[mbarbar]

When collapsing any cycle, we need to potentially handle stores/loads (postProcessNode) and copies/geps (processNode) if the points-to set has changed or if the rep node gets new edges. I originally thought not so for the latter but SCC detection can happen before all nodes in the SCC get the same points-to set.

Adding to worklist/reanalyzing conditionally upon those two conditions requires a lot of refactoring I believe. I've instead added the rep node unconditionally whenever an SCC is collapsed.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 65.82%. Comparing base (ef21931) to head (a465851).

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #1873   +/-   ##
=======================================
  Coverage   65.82%   65.82%           
=======================================
  Files         251      251           
  Lines       24665    24668    +3     
  Branches     4661     4662    +1     
=======================================
+ Hits        16235    16238    +3     
  Misses       8430     8430           

Files with missing lines	Coverage Δ
svf/lib/WPA/Andersen.cpp	93.65% <100.00%> (+0.05%)	⬆️

🚀 New features to boost your workflow:

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

[mbarbar]

Interestingly this solves a real problem but the stack-1.bc stuff we were looking at before has some other problems too it turns out. Investigating still (that is probably the access-from-0-offset vs. access-from-base you mentioned but I want to confirm it).

[yuleisui]

This is an issue orthogonal to the gep offset. Shall I merge this one first?

[mbarbar]

Yes, it should be good to merge.