[llvm] [SLP] Fallback to sext when widening vectorized operands (PR #106726)

[Alexey Bataev via llvm-commits]

alexey-bataev wrote:

It does not fix the issue, just hides it. The problem is that there is discrepancy between getOperandEntry function and vectorizeOperand function. They are returning 2 different entries, which has different minbitwidth processing

https://github.com/llvm/llvm-project/pull/106726