Iter 8a: closure-pair ABI for fn-values

Every fn-value is now a `ptr` to a closure pair `{ ptr thunk, ptr env }`,
not a raw fn-pointer. Top-level fns get an auto-generated adapter
`@ail_<m>_<f>_adapter(ptr %_env, params...)` that ignores env and
forwards to the real fn, plus a static closure constant
`@ail_<m>_<f>_clos = { @adapter, null }`. References to a fn as a
value return the closure-pair address.

Indirect calls now GEP the thunk + env slots, load both, and call
`thunk(env, args...)`. Direct calls to statically-known callees stay
on the original fast path (no adapter).

This is the ABI groundwork for Iter 8b lambdas: a captured-env closure
will reuse the same value shape, just with a non-null env produced by
malloc.

Tests: 48 still green. IR snapshots refreshed (every fn now carries an
adapter + static closure pair). Iter 7's hof.ail.json prints 42
unchanged — `apply(inc, 41)` now hands `@ail_hof_inc_clos` to apply,
which unpacks and indirect-calls as designed.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

crates/ail/tests/snapshots/hello.ll

@@ -8,12 +8,19 @@ declare i32 @printf(ptr, ...)
 declare i32 @puts(ptr)
 declare ptr @malloc(i64)
 
+@ail_hello_main_clos = private unnamed_addr constant { ptr, ptr } { ptr @ail_hello_main_adapter, ptr null }
 define i8 @ail_hello_main() {
 entry:
   call i32 @puts(ptr @.str_hello_str_0)
   ret i8 0
 }
 
+define i8 @ail_hello_main_adapter(ptr %_env) {
+entry:
+  %r = call i8 @ail_hello_main()
+  ret i8 %r
+}
+
 
 define i32 @main() {
   call i8 @ail_hello_main()

crates/ail/tests/snapshots/list.ll

@@ -8,6 +8,8 @@ declare i32 @printf(ptr, ...)
 declare i32 @puts(ptr)
 declare ptr @malloc(i64)
 
+@ail_list_sum_list_clos = private unnamed_addr constant { ptr, ptr } { ptr @ail_list_sum_list_adapter, ptr null }
+@ail_list_main_clos = private unnamed_addr constant { ptr, ptr } { ptr @ail_list_main_adapter, ptr null }
 define i64 @ail_list_sum_list(ptr %arg_xs) {
 entry:
   %v1 = load i64, ptr %arg_xs, align 8
@@ -32,6 +34,12 @@ mjoin.2:
   ret i64 %v9
 }
 
+define i64 @ail_list_sum_list_adapter(ptr %_env, ptr %a0) {
+entry:
+  %r = call i64 @ail_list_sum_list(ptr %a0)
+  ret i64 %r
+}
+
 define i8 @ail_list_main() {
 entry:
   %v1 = call ptr @malloc(i64 8)
@@ -59,6 +67,12 @@ entry:
   ret i8 0
 }
 
+define i8 @ail_list_main_adapter(ptr %_env) {
+entry:
+  %r = call i8 @ail_list_main()
+  ret i8 %r
+}
+
 
 define i32 @main() {
   call i8 @ail_list_main()

crates/ail/tests/snapshots/max3.ll

@@ -8,6 +8,9 @@ declare i32 @printf(ptr, ...)
 declare i32 @puts(ptr)
 declare ptr @malloc(i64)
 
+@ail_max3_max_clos = private unnamed_addr constant { ptr, ptr } { ptr @ail_max3_max_adapter, ptr null }
+@ail_max3_max3_clos = private unnamed_addr constant { ptr, ptr } { ptr @ail_max3_max3_adapter, ptr null }
+@ail_max3_main_clos = private unnamed_addr constant { ptr, ptr } { ptr @ail_max3_main_adapter, ptr null }
 define i64 @ail_max3_max(i64 %arg_a, i64 %arg_b) {
 entry:
   %v1 = icmp sgt i64 %arg_a, %arg_b
@@ -21,6 +24,12 @@ join.2:
   ret i64 %v3
 }
 
+define i64 @ail_max3_max_adapter(ptr %_env, i64 %a0, i64 %a1) {
+entry:
+  %r = call i64 @ail_max3_max(i64 %a0, i64 %a1)
+  ret i64 %r
+}
+
 define i64 @ail_max3_max3(i64 %arg_a, i64 %arg_b, i64 %arg_c) {
 entry:
   %v1 = icmp sgt i64 %arg_a, %arg_b
@@ -50,6 +59,12 @@ join.2:
   ret i64 %v9
 }
 
+define i64 @ail_max3_max3_adapter(ptr %_env, i64 %a0, i64 %a1, i64 %a2) {
+entry:
+  %r = call i64 @ail_max3_max3(i64 %a0, i64 %a1, i64 %a2)
+  ret i64 %r
+}
+
 define i8 @ail_max3_main() {
 entry:
   %v1 = call i64 @ail_max3_max3(i64 3, i64 17, i64 9)
@@ -57,6 +72,12 @@ entry:
   ret i8 0
 }
 
+define i8 @ail_max3_main_adapter(ptr %_env) {
+entry:
+  %r = call i8 @ail_max3_main()
+  ret i8 %r
+}
+
 
 define i32 @main() {
   call i8 @ail_max3_main()

crates/ail/tests/snapshots/sum.ll

@@ -8,6 +8,8 @@ declare i32 @printf(ptr, ...)
 declare i32 @puts(ptr)
 declare ptr @malloc(i64)
 
+@ail_sum_sum_clos = private unnamed_addr constant { ptr, ptr } { ptr @ail_sum_sum_adapter, ptr null }
+@ail_sum_main_clos = private unnamed_addr constant { ptr, ptr } { ptr @ail_sum_main_adapter, ptr null }
 define i64 @ail_sum_sum(i64 %arg_n) {
 entry:
   %v1 = icmp eq i64 %arg_n, 0
@@ -24,6 +26,12 @@ join.2:
   ret i64 %v6
 }
 
+define i64 @ail_sum_sum_adapter(ptr %_env, i64 %a0) {
+entry:
+  %r = call i64 @ail_sum_sum(i64 %a0)
+  ret i64 %r
+}
+
 define i8 @ail_sum_main() {
 entry:
   %v1 = call i64 @ail_sum_sum(i64 10)
@@ -31,6 +39,12 @@ entry:
   ret i8 0
 }
 
+define i8 @ail_sum_main_adapter(ptr %_env) {
+entry:
+  %r = call i8 @ail_sum_main()
+  ret i8 %r
+}
+
 
 define i32 @main() {
   call i8 @ail_sum_main()

crates/ail/tests/snapshots/ws_main.ll

@@ -8,12 +8,20 @@ declare i32 @printf(ptr, ...)
 declare i32 @puts(ptr)
 declare ptr @malloc(i64)
 
+@ail_ws_lib_add_clos = private unnamed_addr constant { ptr, ptr } { ptr @ail_ws_lib_add_adapter, ptr null }
+@ail_ws_main_main_clos = private unnamed_addr constant { ptr, ptr } { ptr @ail_ws_main_main_adapter, ptr null }
 define i64 @ail_ws_lib_add(i64 %arg_a, i64 %arg_b) {
 entry:
   %v1 = add i64 %arg_a, %arg_b
   ret i64 %v1
 }
 
+define i64 @ail_ws_lib_add_adapter(ptr %_env, i64 %a0, i64 %a1) {
+entry:
+  %r = call i64 @ail_ws_lib_add(i64 %a0, i64 %a1)
+  ret i64 %r
+}
+
 define i8 @ail_ws_main_main() {
 entry:
   %v1 = call i64 @ail_ws_lib_add(i64 2, i64 3)
@@ -21,6 +29,12 @@ entry:
   ret i8 0
 }
 
+define i8 @ail_ws_main_main_adapter(ptr %_env) {
+entry:
+  %r = call i8 @ail_ws_main_main()
+  ret i8 %r
+}
+
 
 define i32 @main() {
   call i8 @ail_ws_main_main()