[llvm] VPlan: simplify mul with zero (PR #127521)

[Luke Lau via llvm-commits]

lukel97 wrote:

> See surrounding code. Yes, LV introduces plenty of trivial constructs like mul with 1, not(not(V)) etc. when building the VPlan.

Is there a test for this? The test affected in this PR has a mul of 0 in the input

https://github.com/llvm/llvm-project/pull/127521