fix issue 0143: pack-as-[]Type built as []Any — build it as []type_value

buildPackSliceValue (lower/pack.zig) materialized a bare `$<pack>` []Type slice
as []Any (16-byte elements) — a stale mapping from before the dedicated Type
builtin (.type_value, 8 bytes) replaced Type -> .any. It stored 8-byte const_type
words into 16-byte slots, so a []Type reader (8-byte stride) read [t0, pad, t1, ...]
instead of [t0, t1, ...]. The legacy interp's tagged-Value model hid it; the
byte-accurate comptime VM exposed it (a `..$args` pack forwarded as a []Type
argument across a call read its elements shifted/garbled).

Fix: build the pack-slice array + slice as .type_value (8 bytes). Removed the
stopgap type_name .unresolved guard (379ed05) now that the root cause is fixed.

Regression test examples/0525-packs-pack-as-type-slice-arg.sx (outer(42,"hi",true)
-> inner($args: []Type) -> "i64 string bool"). 700/0 both gates; issue 0143 RESOLVED.

src/ir/comptime_vm.zig

@@ -1500,12 +1500,6 @@ pub const Vm = struct {
                     }
                     return self.failMsg("comptime type_name: arg is not a Type value or an Any box");
                 };
-                // An `.unresolved` TypeId means the read produced a bad type (e.g. a
-                // mis-strided `[]Type` slice over an Any-boxed pack — see issue 0143):
-                // the VM can't faithfully name it, so BAIL rather than emit
-                // "<unresolved>". (The legacy reads Values, not bytes, so it gets the
-                // real type; the fallback then handles this correctly.)
-                if (tid == .unresolved) return self.failMsg("comptime type_name: unresolved type (bad slice/pack read — see issue 0143)");
                 return try self.makeStringValue(table, table.typeName(tid));
             },
             // type_info($T) → reflect a type INTO a TypeInfo VALUE (the inverse of