Modify SimplifyArmIdentity so it can trigger on mir-opt-level=1

[wesleywiser]

I also added test cases to make sure the optimization can fire on all of
these cases:

fn case_1(o: Option<u8>) -> Option<u8> {
  match o {
    Some(u) => Some(u),
    None => None,
  }
}

fn case2(r: Result<u8, i32>) -> Result<u8, i32> {
  match r {
    Ok(u) => Ok(u),
    Err(i) => Err(i),
  }
}

fn case3(r: Result<u8, i32>) -> Result<u8, i32> {
  let u = r?;
  Ok(u)
}

Without MIR inlining, this still does not completely optimize away the
? operator because the Try::into_result(), From::from() and
Try::from_error() calls still exist. This does move us a bit closer to
that goal though because:

• We can now run the pass on mir-opt-level=1

• We no longer depend on the copy propagation pass running which is
  unlikely to stabilize anytime soon.

Fixes #66855

[wesleywiser]

@bors try @rust-timer queue

[rust-timer]

Awaiting bors try build completion

[bors]

⌛ Trying commit b5238fb with merge 69665ae...

[bors]

☀️ Try build successful - checks-azure
Build commit: 69665ae (69665ae05cc719832de28428bece87b264ef69ad)

[rust-timer]

Queued 69665ae with parent b818ccc, future comparison URL.

[rust-timer]

Finished benchmarking try commit 69665ae, comparison URL.

[wesleywiser]

@bors try @rust-timer queue

[rust-timer]

Awaiting bors try build completion

[bors]

⌛ Trying commit f0d5e44 with merge be3cc06...

[bors]

☀️ Try build successful - checks-azure
Build commit: be3cc06 (be3cc06352a0c63fce9b16c057a58b7f238051ee)

[rust-timer]

Queued be3cc06 with parent 2cb0b85, future comparison URL.

[rust-timer]

Finished benchmarking try commit be3cc06, comparison URL.

[wesleywiser]

I've manually verified the syn-opt benchmark and this optimization doesn't fire at all (the MIR and LLVM-IR before and after this change is the same) so the "regression" is just noise.

[wesleywiser]

r? @oli-obk

[wesleywiser]

@oli-obk The build failure is legit. I've got a repro of the issue with liballoc and I'll include a regression test when I have a fix.

[bors]

⌛ Testing commit 6de6d70 with merge c892c7cd3cf05930ed4d1fdc7ff025c38dad1969...

[bors]

💔 Test failed - checks-actions

[wesleywiser]

@bors retry

Spurious macOS failure.

[bors]

⌛ Testing commit 6de6d70 with merge 618120aeea8892e23b8b466bab4c1f308cfaae42...

[Dylan-DPC-zz]

@bors retry yield

[bors]

⌛ Testing commit 6de6d70 with merge c6b39cc0035a644fe95217a97de42d261a53d7c0...

[bors]

💔 Test failed - checks-actions

[oli-obk]

@bors retry artifact upload failure

[bors]

⌛ Testing commit 6de6d70 with merge 7c34d8d...

[bors]

☀️ Test successful - checks-actions, checks-azure
Approved by: oli-obk
Pushing 7c34d8d to master...

[nnethercote]

This caused a perf regression for issue-58319. It's not important because that's an artificial microbenchmark that once caused stack overflows and is present just to make sure that doesn't happen again, but I mention it for completeness.

[oli-obk]

Curious that the regression didn't show up in the perf runs in this PR. Maybe some other change snuck in between that caused this change to become a perf regression?

[oli-obk]

Also, according to https://perf.rust-lang.org/detailed-query.html?commit=7c34d8d6629506a596215886e5fc4bb2b04b00ae&base_commit=23ffeea307c31f0c20ebb5a15d5171e0c414629d&benchmark=issue-58319-check&run_name=full the regression is entirely inside borrowck, which isn't affected by this change beyond this change affecting the built rustc. So... did we make rustc slower (I honestly fail to see how)

[wesleywiser]

Total time for the issue-58319 test is in the hundredths of seconds. That may just be too small a runtime for the self-profile time column to be comparable. For the mir_borrowck query, the variance is 1/1000th of a second.