[mlir][tosa] Fix tosa::Select to linalg::generic indexingMaps bug

[AmosLewis]

Summary

The tosa to linalg generic ops indexingMaps rank use is wrong.

	 Find this bug in gpt2 pytorch model lowering to tosa.
	 issue link is here https://github.com/llvm/llvm-project/issues/58154

[rsuderman]

Could you include a test case that demonstrates the failure?

[AmosLewis]

[mlir][tosa] Fix tosa::Select to linalg::generic indexingMaps bug

[AmosLewis]

In D135343#3840381, @rsuderman wrote:

Could you include a test case that demonstrates the failure?

Done

[jpienaar]

LG, we can simplify the test a bit. I'll do that during landing.

mlir/lib/Conversion/TosaToLinalg/TosaToLinalg.cpp

for (Value operand : operation->getOperands()) {
     if (newShape.size() != rank) {
       operand = rewriter.create<tosa::ReshapeOp>(
           loc, RankedTensorType::get(newShape, type.getElementType()), operand,
           rewriter.getI64ArrayAttr(newShape));
     }
 
     operands.push_back(operand);
     indexingMaps.push_back(AffineMap::get(
-        /*dimCount=*/type.getRank(), /*symbolCount=*/0, affineExprs,
+        /*dimCount=*/rank, /*symbolCount=*/0, affineExprs,
         rewriter.getContext()));
   }
 
   indexingMaps.append(operation->getNumResults(),
                       rewriter.getMultiDimIdentityMap(rank));
 
   bool didEncounterError = false;
   auto linalgOp = rewriter.create<linalg::GenericOp>(