#900, #902

Two pre-existing check-green-then-codegen-broken bugs in the same
Unit-as-a-degenerate-component family (both on release/v0.1.0 and main).

#900 — a USER forall<T> function whose body READS @t, inferred to
T = Unit, passed vera check but crashed in codegen: Unit is 0 bytes with
no WASM representation (spec 11.2.2 / 11.3.1), so the monomorphized body's
@T.n slot read lowers to a local.get with no local — the dangling-slot
invariant, whose symptom was hash-seed-dependent (E699, a dropped main,
or a CodegenInvariantError "file a bug report" crash). The checker's
type-argument inference now rejects a body-read forall parameter inferred
to bare Unit with the new E206, per DESIGN.md principle 1 (checkability
over correctness).

The check is NARROWED to bodies that materialize @t (a @T.n SlotRef read,
computed once at registration): a @t parameter the body never reads erases
cleanly from the ABI, so firstInt(@t, @int){ @Int.0 } and ignore(@t){ 0 }
still run at T = Unit and are NOT rejected. Scoped to user-declared
functions the monomorphizer clones — built-in generics
(async/await/collection + prelude combinators) have hand-written codegen
and are excluded, so async(IO.print(...)) keeps its Future<Unit> W002
path. Keyed to bare Unit only: a boxed Option<Unit> stays valid. Covers
every @T-reading arg-pinned entry path (Unit-returning fn, () literal, one
of several params, reads nested in let/if/match).

#902 — a Tuple (or any constructor) with a Unit component (Tuple((), 3)
returning Tuple<Unit, Int> passed vera check but crashed in codegen with
unknown func: failed to find name : _translate_constructor_call
could not infer a WASM type for the Unit field, raised a CodegenSkip, and
dropped the whole function while its caller still called it. A Unit field
is now laid out as a genuine zero-size field (its arg instructions run for
side effects, but it occupies no bytes and stores nothing), mirroring the
existing Unit skip in _translate_let_destruct. Unit detection uses
_is_void_expr (the canonical produces no stack value check), so a
Unit-RETURNING call in field position — IO.print(x), a user @Unit fn, a
void effect op, a ModuleCall to a @Unit fn — is also laid out zero-size
with its side effect preserved, not just a bare () literal. Covers Unit in
any tuple position, multiple Unit fields, nested tuples, a user ADT with a
Unit field, and a side-effecting Unit-returning field.

Adds a check-level negative conformance fixture
(ch02_generic_over_unit_rejected, E206) and a run-level positive one
(ch02_tuple_unit_component), reconciles doc counts (127 conformance
programs), and restores KNOWN_ISSUES to No known bugs.

Co-Authored-By: Claude <[email protected]>
EOF
)