[clang] cmp,1 + je instead of tst, jne

The following code

```c++
#include
#include

void asymmetric_thread_fence_light() { asm volatile("" : : : "memory"); }

struct Reader1 {
void exit();

void notify();

std::atomic rcu_count;
std::uint32_t nested_readers;
std::atomic waiting;
};

void Reader1::exit() {
if (nested_readers) [[unlikely]] {
--nested_readers;
return;
}

asymmetric_thread_fence_light();
rcu_count.store(0, std::memory_order_relaxed);
asymmetric_thread_fence_light();

if (waiting.load(std::memory_order_relaxed)) [[unlikely]] {
notify();
}
}
```

https://godbolt.org/z/7P46qfYY6

In the check
`if (waiting.load(std::memory_order_relaxed)) [[unlikely]] {`
generates `cmp al, 1`

If we invert the condition we get `test al, al`

gcc generates `tst al, al` in both cases.

While I don't know for a fact here, I thought `tst al, al` would be the preferrable codegen.

Reproduce the C++ example from the issue in the linked Compiler Explorer and compare Clang's assembly for the original and inverted conditions with GCC's output. Start by determining why Clang emits cmp against 1 for the atomic bool load; done means establishing whether test-style code generation is correct and, if so, adding the relevant regression coverage.

Written by the indexing model from the issue text.