Abstract interpretation based analysis

[yamaguchi1024]

(this comment is outdated. needs update)

This PR:

• src/exo/dataflow.py implements abstract interpretation based scalar propagation
• Convert all the checks to use DataflowIR based 1) control values 2) abstract values. I still rely on stmts_effs and PostEnv to extract effects from LoopIR, but I don't use them for propagating values.
• Only actual configuration states value sensitive analysis is Check_DeleteConfigWrite. Other checks can actually be strickter, by being value-insensitive. For instance, reordering Config.Load = 1.0 ; Config.Load = 1.0 is invalid, but deleting one of the configurations should be valid.

What I hope to do after merging this PR:

• performance improvement
  • Integrate DataflowIR into LoopIR. DataflowIR is similar to LoopIR so it won't be too much of a hussle
  • Iterate on the Cursor not the tree
  • Cache abstract values on LoopIR
  • Merge controls into the abstract values (integrate ScalarPropagation with GetControlPredicates)
• Precision improvement
  • Support loop-over-arrays

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 97.31622% with 23 lines in your changes missing coverage. Please review.

Project coverage is 88.45%. Comparing base (0cf7e19) to head (dd00c56).

Files	Patch %	Lines
src/exo/dataflow.py	96.23%	19 Missing ⚠️
src/exo/dataflow_pprint.py	97.24%	3 Missing ⚠️
tests/test_dataflow.py	98.66%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #578      +/-   ##
==========================================
+ Coverage   87.74%   88.45%   +0.71%     
==========================================
  Files          84       87       +3     
  Lines       20658    21256     +598     
==========================================
+ Hits        18126    18802     +676     
+ Misses       2532     2454      -78     

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