fix: Handle multiple entry points for Brillig call stack resolution after metadata deduplication

[vezenovm]

Description

Problem*

Resolves

Issue created by #5696.

Summary*

After the creation of the brillig locations map in #5696. we do the following to attach Brillig locations to our generated acir:

        if inserted_func_before {
            return;
        }

        for (brillig_index, call_stack) in generated_brillig.locations.iter() {
            self.brillig_locations.entry(brillig_function_index).or_default().insert(
                OpcodeLocation::Brillig {
                    acir_index: self.opcodes.len() - 1,
                    brillig_index: *brillig_index,
                },
                call_stack.clone(),
            );
        }

This inserts the ACIR index at the location the brillig call was pushed. However, this is only valid for the first Brillig call. Any later Brillig calls will also use this acir index to resolve against the debug metadata.

We just need to avoid trying to match against this acir index as the Brillig call stacks do not require the acir index in order to be resolved against the debug metadata.

I also fixed the flamegraph in this PR.

For the following snippet where x == 0 causing a failure:

fn main(x: Field) {
    unsafe {
        assert(1 == conditional_wrapper(!x as bool));
        assert(1 == conditional_wrapper(x as bool));
    }
}
unconstrained fn conditional_wrapper(x: bool) -> Field {
    conditional(x)
}
unconstrained fn conditional(x: bool) -> Field {
    assert(x);
    1
}

Before:

After:

The flame graph also is operating as expected now. Here some example output from one of the Brillig entry point of the brillig_calls test:

Additional Context

The assert messages still use the Brillig variant of OpcodeLocation. This is a breaking serialization change so I have opted for not changing that here and making a separate struct type for Brillig opcode locations.

Documentation*

Check one:

• No documentation needed.
• Documentation included in this PR.
• [For Experimental Features] Documentation to be submitted in a separate PR.

PR Checklist*

• I have tested the changes locally.
• I have formatted the changes with Prettier and/or cargo fmt on default settings.

[github-actions]

Changes to Brillig bytecode sizes

Generated at commit: 2bb9ca2423fa0c9f39fd3356668b0f0974c9fb77, compared to commit: 335de054dfcda366df50cc215900910ebdc8be63

There are no changes in circuit sizes

[michaeljklein]

LGTM

[TomAFrench]

LGTM, I think we have quite a bit of tech debt surrounding debug info so it would probably make sense to sit down, take stock of it and see if we can simplify how we handle it.