Couple (maybe?) missing optimizations for suspend ...

# compiler

[Youssef Shoaib [MOD]]

Couple (maybe?) missing optimizations for suspend lambdas on JVM:
• it seems that TCO is not applied to them; e.g. I'm seeing bytecode that decompiles to:

case <some-label-number>: {
    ResultKt.throwOnFailure((Object)$result);
    Object object2 = $result;
    return object2;
}

• it seems that their closures are never spilled out (in fact, they have their closed-over values as final fields)
Is this intentional design, or just overlooked corner cases?
(LMK if this fits better in #C1CFAFJSK, but I figured the low-level compilation details are better discussed here)