[llvm] [DFAJumpThreading] Enable DFAJumpThread by default. (PR #157646)

[Hongyu Chen via llvm-commits]

XChy wrote:

I profiled libclamav_nsis_LZMADecode with valgrind. I found that the SLPVectorizer becomes slower after DFAJumpThreading. Specifically, the time proportion of `SLPVectorizer::BoUpSLP::buildTreeRec` increases from 10% to 37%.

Before enableing DFAJumpThreading:
<img width="849" height="674" alt="image" src="https://github.com/user-attachments/assets/5c83241c-14cd-4ac7-bc1d-9223e5cb09fc" />

After:
<img width="849" height="674" alt="image" src="https://github.com/user-attachments/assets/9a8e3010-fe7d-429f-b51b-409805c2f743" />


https://github.com/llvm/llvm-project/pull/157646