huge refactor for high RewritePatterns usage and less ad-hoc cpp code

remove Spatial many ops in favor of tensor ops like in pim

src/PIM/Pass/PimCodegen/HostConstantFolding/Patterns/Subview.cpp

@@ -128,6 +128,7 @@ static LogicalResult rewriteSubviewCopyLikeOp(CopyOp copyOp,
   return success();
 }
 
+// Splits core copies through subviews into contiguous copy chunks for codegen.
 struct RewriteCoreSubviewCopyPattern final : OpRewritePattern<pim::PimMemCopyOp> {
   using OpRewritePattern::OpRewritePattern;
 
@@ -162,6 +163,7 @@ struct RewriteCoreSubviewCopyPattern final : OpRewritePattern<pim::PimMemCopyOp>
   }
 };
 
+// Splits host-to-device subview loads into contiguous copy chunks.
 struct RewriteHostSubviewLoadPattern final : OpRewritePattern<pim::PimMemCopyHostToDevOp> {
   using OpRewritePattern::OpRewritePattern;
 
@@ -193,6 +195,7 @@ struct RewriteHostSubviewLoadPattern final : OpRewritePattern<pim::PimMemCopyHos
   }
 };
 
+// Splits device-to-host subview stores into contiguous copy chunks.
 struct RewriteHostSubviewStorePattern final : OpRewritePattern<pim::PimMemCopyDevToHostOp> {
   using OpRewritePattern::OpRewritePattern;
 
@@ -224,6 +227,7 @@ struct RewriteHostSubviewStorePattern final : OpRewritePattern<pim::PimMemCopyDe
   }
 };
 
+// Folds constant subviews used as core weights into standalone globals.
 struct FoldConstantCoreSubviewPattern final : OpRewritePattern<memref::SubViewOp> {
   using OpRewritePattern::OpRewritePattern;