[llvm] [CodeGen] Compact cost table entries (PR #202836)

[Alexis Engelke via llvm-commits]

aengelke wrote:

Looks like this slightly regresses performance, I expect due to indexing into 6B structs is less efficient than indexing into 8B structs: https://llvm-compile-time-tracker.com/compare.php?from=3fbe826e3707540c9b7f942c573fbcb2d51b6bcc&to=62708d1030229006b70bf1b5a46dbb56d2d058ec&stat=instructions:u

https://github.com/llvm/llvm-project/pull/202836