[llvm] [TailDuplicator] Add a limit on the size of predecessors (PR #78582)

[Kai Luo via llvm-commits]

bzEq wrote:

The cause is early-tailduplication transforms normal `switch` to computed-gotos, which means a lot of indirect branches are generated, adding quadratic numbers of edges inside the these cases, that's to say at the end of each case, the control flow is able to jump to other cases directly. As a consequence, passes run after `early-tailuplication` get lag due to a more complex CFG.

https://github.com/llvm/llvm-project/pull/78582