Profiler probes for non-sequence steps

[cameel]

I realized that the only steps covered by our profiler are the ones that go through runSequence(). Disambiguator, NameSimplifier, StackCompressor, StackLimitEvader, NameSimplifier, VarNameCleaner are not included, which undercounts the actual impact of the optimizer by quite a bit (33% in case of Uniswap). This PR adds probes for these elements.

[cameel]

Here's how it looks in benchmarks. The tables show the timing of the added probes (in italics), with total optimization time and top 3 steps included for comparison.

Looks like it's closer to 15%. I assumed 33% based on the fact that for Uniswap I used to see ~60 s of optimization time and now that's up to ~80 s, but apparently this run was just particularly slow.

The impact of NameSimplifier being 4-6% is a lot more than I expected.

OpenZeppelin

Compilation time: 33 s.
Time spent in added probes: 4 s (16% of optimization time).

Time %	Time	Calls	Scope
0.1%	0.031 s	274	ConstantOptimiser
0.1%	0.033 s	274	VarNameCleaner
1.1%	0.261 s	274	Disambiguator
3.8%	0.890 s	274	StackLimitEvader
4.3%	0.998 s	274	StackCompressor
6.3%	1.470 s	548	NameSimplifier
7.5%	1.750 s	2740	UnusedPruner
7.9%	1.834 s	1370	UnusedAssignEliminator
16.8%	3.917 s	3294	CommonSubexpressionEliminator
100.0%	23.259 s	40588	TOTAL

Uniswap

Compilation time: 130 s.
Time spent in added probes: 12 s (15% of optimization time).

Time %	Time	Calls	Scope
0.1%	0.051 s	270	ConstantOptimiser
0.3%	0.261 s	270	VarNameCleaner
0.8%	0.612 s	270	Disambiguator
4.4%	3.508 s	540	NameSimplifier
4.5%	3.575 s	270	StackLimitEvader
4.9%	3.965 s	270	StackCompressor
7.6%	6.086 s	1350	UnusedAssignEliminator
8.2%	6.570 s	1391	ExpressionSimplifier
17.2%	13.785 s	3281	CommonSubexpressionEliminator
100.0%	80.193 s	40206	TOTAL

Eigenlayer

Compilation time: 556 s.
Time spent in added probes: 48 s (15% of optimization time).

Time %	Time	Calls	Scope
0.1%	0.272 s	376	ConstantOptimiser
0.6%	1.875 s	376	VarNameCleaner
0.7%	2.163 s	376	Disambiguator
3.5%	11.434 s	752	NameSimplifier
4.8%	15.680 s	376	StackLimitEvader
5.1%	16.761 s	376	StackCompressor
7.9%	26.112 s	1972	ExpressionSimplifier
9.3%	30.850 s	3852	LiteralRematerialiser
17.0%	56.100 s	4604	CommonSubexpressionEliminator
100.0%	329.952 s	56200	TOTAL

[cameel]

I wonder why we're running NameSimplifier twice. It has been that way ever since it was added in #9591. I don't see any specific justification for this in that PR though.

Removing it could shave off 2-3% of the optimizer running time.

[clonker]

I was thinking that perhaps it yields nicer names in some cases, but at least for our test suite it doesn't even change anything.

[cameel]

Nice. You could submit that as a PR then. We should run it by Daniel just to make sure there wasn't any obscure reason he's aware of but if it's having no effect, we should really get rid of it. It's too expensive to run it without a good reason.