fix(pxe): prevent contract sync deadlock on nested syncs

[nchamo]

Why

The contract sync service uses a semaphore to limit how many scope syncs run concurrently. The problem: a scope sync can trigger a nested sync (a cross-contract utility call from sync_state) that needs a slot of its own. So when 5 syncs each hold a slot and each waits on a nested sync, no slots are left for those nested syncs, and everything deadlocks. And because the semaphore was shared across the whole PXE, unrelated transactions could fill it and trigger the same deadlock.

Fix

Give each sync call its own semaphore, rate-limiting only the scopes within that single call. Those scopes don't depend on each other, so bounding them can't deadlock. A nested sync builds its own limiter, so it always has a slot and can make progress. We still cap resource usage when a single call has many scopes.

[AztecBot]

Flakey Tests

🤖 says: This CI run detected 1 tests that failed, but were tolerated due to a .test_patterns.yml entry.

\033FLAKED\033 (8;;http://ci.aztec-labs.com/d1a3d99f7f8f3ac8�d1a3d99f7f8f3ac88;;�): yarn-project/scripts/run_test.sh bb-prover/src/avm_proving_tests/avm_check_circuit_token.test.ts (5s) (code: 0)

[mverzilli]

Oof, very nice catch. How did you stumble upon this?

[nchamo]

I got the deadlock here: #23938