Use `BLSR` fusions to break dependency chains even when intermediate expression value is consumed

[Godbolt](https://c.godbo.lt/#g:!((g:!((g:!((h:codeEditor,i:(filename:'1',fontScale:14,fontUsePx:'0',j:1,lang:___c,selection:(endColumn:1,endLineNumber:7,positionColumn:1,positionLineNumber:7,selectionStartColumn:1,selectionStartLineNumber:1,startColumn:1,startLineNumber:1),source:'%23include+%3Cstdint.h%3E%0A%0Auint64_t+foo(uint64_t+z)+%7B%0A++++uint64_t+w+%3D+z+-+1%3B%0A++++return+(z+%26+w)+*+w%3B%0A%7D%0A'),l:'5',n:'0',o:'C+source+%231',t:'0')),k:47.14192766144739,l:'4',n:'0',o:'',s:0,t:'0'),(g:!((h:compiler,i:(compiler:cclang_trunk,filters:(b:'0',binary:'1',binaryObject:'1',commentOnly:'0',debugCalls:'1',demangle:'0',directives:'0',execute:'1',intel:'0',libraryCode:'0',trim:'1',verboseDemangling:'0'),flagsViewOpen:'1',fontScale:14,fontUsePx:'0',j:1,lang:___c,libs:!(),options:'-O3+--target%3Dx86_64-linux+-march%3Dznver5',overrides:!(),selection:(endColumn:1,endLineNumber:1,positionColumn:1,positionLineNumber:1,selectionStartColumn:1,selectionStartLineNumber:1,startColumn:1,startLineNumber:1),source:1),l:'5',n:'0',o:'+x86-64+clang+(trunk)+(Editor+%231)',t:'0')),k:52.858072338552596,l:'4',m:100,n:'0',o:'',s:0,t:'0')),l:'2',n:'0',o:'',t:'0')),version:4)

```c
#include

uint64_t foo(uint64_t z) {
uint64_t w = z - 1;
return (z & w) * w;
}
```

Emits:

```asm
foo:
lea rax, [rdi - 1]
and rdi, rax ; can't be computed in parallel with `lea`
imul rax, rdi
ret
```

I was expecting:

```asm
foo:
lea rax, [rdi - 1]
blsr rdi, rdi ; can be computed in parallel with `lea`
imul rax, rdi
ret
```

Start with the linked Godbolt C reproducer and inspect the LLVM x86 code-generation path responsible for selecting BLSR for the `(z & (z - 1))` expression. Compare the emitted assembly with the expected sequence, and consider the work complete when the intermediate value remains usable while the BLSR operation can be scheduled in parallel with the subtraction.

Written by the indexing model from the issue text.