Allow explicit cast of an expression to its own type

[aeron-gh]

Closes #5527.

The spec (section 8.12.1) says an explicit cast of an expression to its own type is a no-op and is allowed. The type checker did not honour this for struct/header/union types: a cast to such a type was only accepted when the operand was a struct initializer ((S){...}), otherwise it reported cast not supported. So a redundant self-cast such as

struct s { bit t; bit t1; }

s func1(s t1) {
    return (s)(s)(s)t1;
}

failed to type-check even though every cast targets t1's own type.

postorder(IR::Cast) now checks, when the destination is a struct-like type, whether the source and destination types are equivalent, and if so treats the cast as a no-op. The existing struct-initializer and invalid-header paths are untouched, so the only programs that change behaviour are self-casts that previously errored — the redundant casts are simply dropped (e.g. the program above pretty-prints back to return t1;).

testdata/issue3233.p4 was an error test for exactly this case; it now compiles, so it moves from p4_16_errors to p4_16_samples with the generated outputs.

I confirmed all existing explicit-cast-* / struct / union / tuple cast error tests still fail with their expected diagnostics (the change is gated on type equivalence, so genuinely-incompatible casts are unaffected).

[jafingerhut]

I reviewed only the updated test case and expected output files for them. Those look good to me. I have not attempted to review the C++ code changes in the compiler, leaving that part of the review for others more knowledgeable about compiler internals.