[TIR] Introduce ReduceBranchingThroughOvercompute

[Lunderberg]

This PR introduces a new transform, which attempts to reduce branching by adding overcompute, where the overcompute has provably no effect on any observable value outside the PrimFunc.

[tvm-bot]

Thanks for contributing to TVM! Please refer to the contributing guidelines https://tvm.apache.org/docs/contribute/ for useful information and tips. Please request code reviews from Reviewers by @-ing them in a comment.

• cc @Hzfengsy, @junrushao _{See #10317 for details}
• Built docs for commit 7d94147 can be found here.

_{Generated by tvm-bot}

[Lunderberg]

@csullivan This PR is ready for review and passing all unit tests, if you have a chance to review.

[csullivan]

LGTM, one small removal and this looks good to go.

[csullivan]

@tvm-bot rerun

[LeiWang1999]

hi @Lunderberg , did we implement Stmt Simplify(Stmt stmt, arith::Analyzer* analyzer);?

[Lunderberg]

@LeiWang1999 We did, yes. It's located here. Edit: Disregard that. You are right in your comment on #16560, there wasn't actually an implementation added for the prototype of Stmt Simplify.

That said, in most cases it's better to apply tir::transform::Simplify instead. That way, the full context is available for a single simplification call. Simplifying a statement in isolation may miss context-dependent simplifications. (e.g. Within an if i==0 block, if i + 1 < 16 could be simplified to to True.)