Turning on clang's noundef analysis flag by default

Hello all,

As described in the mail that I’ve sent in August, we’d like to turn on clang’s enable-noundef-analysis by default to address the performance regression after some miscompilation bugs are fixed.
(The bug fixes are not landed yet; waiting for slowdowns to be handled. The prototype of bug fix is https://reviews.llvm.org/D106041)

The two patches are: https://reviews.llvm.org/D105169 , https://reviews.llvm.org/D108453 .
These patches help LLVM analyze that variables are well-defined, facilitating optimizations that are correct only if vars are neither undef nor poison.
They are splitted for readability, but they will be merged into one commit and pushed if accepted.

The first link has the numbers about performance improvements after turning the flag on.

Any questions or concerns about enabling the flag are appreciated.

One major concern that we’ve received was that enabling the flag could raise merge conflicts and unit test failures in the downstream projects.
To handle this issue, we will disable the noundef flag for requested tests.
For any requested tests, we will add -disable-noundef-analysis to the flags at // RUN: %clang_cc1 ... to avoid changes in CHECKs. Please feel free to let us know.