[llvm] [AggressiveInstCombine] Add shufflevector support to TruncInstCombine (PR #220218)

[Simon Pilgrim via llvm-commits]

https://github.com/RKSimon commented:

No strong objections - this is verging on needing to be const-driven, but we could argue that for any vector type ops folded through trunc/ext casts (e.g. x86 vXi16 MUL is much cheaper than vXi64). And its a canonicalization most backends can work with.

https://github.com/llvm/llvm-project/pull/220218