transformations: do not insert affine.apply ops when streaming

[superlopuh]

We encode the affine map constraints in the streaming region setup, so we don't need the results of the affine.apply operation, unlike in the memref.load case. We could dce them away, but it feels better to not insert them in the first place.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 89.80%. Comparing base (f33372f) to head (8896dab).
Report is 1 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #2758   +/-   ##
=======================================
  Coverage   89.80%   89.80%           
=======================================
  Files         372      372           
  Lines       47805    47815   +10     
  Branches     7328     7330    +2     
=======================================
+ Hits        42930    42940   +10     
  Misses       3737     3737           
  Partials     1138     1138           

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

[PapyChacal]

I'm a big fan of "just DCE it later" myself when it just works, but LGTM!

NB: You can also just manually reuse DCE from your pass, i.e., don't complexify your pattern and still get your pass outputting simpler IR. Just mentioning the idea, again, this LGTM 🙂

[superlopuh]

I completely agree, hence this PR :)