aarch64: Atomic umin/umax does not zero-extend properly

[alexcrichton]

This input:

test run
target aarch64

function %rmw_umin_i8(i32) -> i8 {
    ss0 = explicit_slot 8
block0(v0: i32):
    v1 = stack_addr.i64 ss0
    v2 = iconst.i8 32
    store.i8 v2, v1
    v3 = ireduce.i8 v0
    v4 = atomic_rmw.i8 umin v1, v3
    v5 = load.i8 v1
    return v5
}
; run: %rmw_umin_i8(0x110) == 16

yields:

$ cargo run -p cranelift-tools --target aarch64-unknown-linux-gnu test case_rmw_umin_i8.clif
    Finished `dev` profile [unoptimized + debuginfo] target(s) in 0.10s
     Running `qemu-aarch64 -L /usr/aarch64-linux-gnu -E LD_LIBRARY_PATH=/usr/aarch64-linux-gnu/lib -E WASMTIME_TEST_NO_HOG_MEMORY=1 /home/alex/code/wasmtime2/target/aarch64-unknown-linux-gnu/debug/clif-util test case_rmw_umin_i8.clif`
[2026-07-25T00:04:36Z ERROR cranelift_filetests::concurrent] FAIL: run
FAIL case_rmw_umin_i8.clif: run

Caused by:
    Failed test: run: %rmw_umin_i8(272) == 16, actual: 32
1 tests
Error: 1 failure

This only affects when has_lse is disabled, and notably the lowering of U{min,max} are missing zero-extensions of loads. And LLM writeup, if helpful, is here