chore(test): Add smoke test for AST generation

[aakoshh]

Description

Problem*

Resolves #7883

Summary*

Add a smoke test to the noir_ast_fuzzer which generates random ASTs and executes them with random inputs, without comparing the results against any oracle, just asserting that the compiler didn't crash. This should help catch problems with the AST or input generation. The testing budget is limited to 10s, so this is not a replacement for cargo fuzz, just a property based test.

Changes function call generation so that we don't call ACIR functions from Brillig functions. In theory there is nothing wrong with doing so, but I just realised debugging an inlining crash that in the normal compiler flow the Monomorphizer would duplicate ACIR functions called from Brillig as unconstrained, which I have not done; in the generated AST so far it was allowed to call ACIR from Brillig. In inlining the recursions weren't properly recognised there. We could duplicate them, but that would make the programs bigger and harder to deal with, and skew frequencies. Instead by not allowing such calls, we pretend that we have already done it. It can result in ACIR functions never being called if main is unconstrained.

Additional Context

To execute the test run the following command:

cargo test -p noir_ast_fuzzer --test smoke

When the test fails, it prints a seed. The failure can be debugged by putting that seed in the NOIR_ARBTEST_SEED env var:

NOIR_ARBTEST_SEED=0x1796975f00100000 cargo test -p noir_ast_fuzzer --test smoke

Currently it fails with various seeds:

• 0x7d5c8fc800004c4b: has open ticket Compiler crashes unrolling for loop with negative boundaries #8009
• 0x9ec585d200002710: simplify_cfg.rs:224:69: index out of bounds: the len is 2 but the index is 113 should have been fixed by Compiler crashes with --force-brillig, passes with ACIR #7964
• 0x644ccb3700100000: created Compiler crash in brillig_globals: function has multiple entry points #8069
• 0x4b01adb300001400: probably Compiler crash involving comparing u128 cast from Field with a condition #8024
• 0xb002524f00001900: created Compiler crash in unrolling: Expected to find the induction variable by end of loop iteration #8073
• 0xf23adcb000100000: inlining.rs:359:17: internal error: entered unreachable code: All Value::Params should already be known from previous calls to translate_block. Unknown value v22 = Param { block: Id(1), position: 0, typ: Numeric(Unsigned { bit_size: 8 }) } -> Fixed by not calling ACIR from Brillig
• 0x24dcefb700100000: inlining.rs:280:13: Attempted to recur more than 1000 times during inlining function 'func_5' -> Fixed by not calling ACIR from Brillig
• 0xf2c25eab00100000: created Compiler crash in acir_context: attempt to divide by zero #8175
• 0xbb5e74bd00100000: created Compiler crash in simplify.rs: index out of bounds #8174
• 0xfbed207000100000: created Compiler crash in acir_context: attempted to divide by constant larger than operand type #8195
• 0x17ed7a1f00100000: created Compiler crash in brillig_ir/instructions: Constant 2 does not fit in bit size 1 #8198
• 0x14fcce9800100000: created Compiler crash in simplify.rs: index out of bounds: Part 2 #8199

These seeds might not work any more, because any change we make to the generation will result in a different program, but the tickets have the minimised version of the code that caused the problem.

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.