Write more testcases for the fnc_tree parser

[ZuseZ4]

Those get handled incorrectly rn:

#[no_mangle]
fn myFn(a: [f64; 1]) { unimplemented!() }

#[no_mangle]
fn myBn(a: (f64, i32)) { unimplemented!() }

These got fixed in the meantime. Probably worth adding to rustc as regression tests.

#[no_mangle]
fn myDn(a: [f64; 2]) { unimplemented!() }

#[no_mangle]
fn mygn(a: (f64, f64, f64)) { unimplemented!() }

[ZuseZ4]

Also it would be lovely to just fuzz-test this parser, it should be easy enough to get it somewhat standalone (and it's probably the most correctness critical piece of Rust-Enzyme).