Allow early otherwise branch on mir-opt-level=2

[spastorino]

Let's test the performance of making this change, note that mir-opt-level is 2 by default in optimized builds.

r? @oli-obk

[spastorino]

@bors try @rust-timer queue

[rust-timer]

Awaiting bors try build completion.

@rustbot label: +S-waiting-on-perf

[bors]

⌛ Trying commit f37adef with merge 76689995512a02828b62e1adda15545c90423f80...

[bors]

☀️ Try build successful - checks-actions
Build commit: 76689995512a02828b62e1adda15545c90423f80 (76689995512a02828b62e1adda15545c90423f80)

[rust-timer]

Queued 76689995512a02828b62e1adda15545c90423f80 with parent 1d6b0f6, future comparison URL.

[rust-timer]

Finished benchmarking try commit (76689995512a02828b62e1adda15545c90423f80): comparison url.

Benchmarking this pull request likely means that it is perf-sensitive, so we're automatically marking it as not fit for rolling up. Please note that if the perf results are neutral, you should likely undo the rollup=never given below by specifying rollup- to bors.

Importantly, though, if the results of this run are non-neutral do not roll this PR up -- it will mask other regressions or improvements in the roll up.

@bors rollup=never
@rustbot label: +S-waiting-on-review -S-waiting-on-perf

[bjorn3]

Many improvements of up to 0.6%. Some regressions up to 0.5%.

[oli-obk]

Any changes to non *-opt runs must either be noise or changes in how this optimization makes rustc faster/slower.

cc @rust-lang/wg-mir-opt I think we have our first candidate for a MIR optimization that we only turn on in release mode.

As to what this optimization does, I think the documentation explains that best:

rust/compiler/rustc_mir/src/transform/early_otherwise_branch.rs

Lines 8 to 23 in f37adef

	/// This pass optimizes something like
	/// ```text
	/// let x: Option<()>;
	/// let y: Option<()>;
	/// match (x,y) {
	/// (Some(_), Some(_)) => {0},
	/// _ => {1}
	/// }
	/// ```
	/// into something like
	/// ```text
	/// let x: Option<()>;
	/// let y: Option<()>;
	/// let discriminant_x = // get discriminant of x
	/// let discriminant_y = // get discriminant of y
	/// if discriminant_x != discriminant_y || discriminant_x == None {1} else {0}

[tmiasko]

I think the transformation is incorrect for reasons described in #78496 (comment).

[oli-obk]

@spastorino please add a check for unsound mir opts to this optimization, referencing the linked issue.

[spastorino]

Closing this one as I've just opened #83277.