make /server/status/ healthcheck endpoints non-blocking with respect to the compiler pool

[zackelan]

current behavior:

calling the the /server/status/alive and/or /server/status/ready healthcheck endpoints makes an RPC to the compiler pool, and if the system is overly busy, that RPC may wait in a queue before it's serviced, potentially causing the healthcheck to time out

desired behavior:

make the check non-blocking so that it won't time out when the system is under heavy load

sketch of a design:

• keep a timestamp of the most recent successful RPC call made to the compiler (time.perf_counter_ns is fine for this, because we don't care about calendar time, just elapsed time)

• in the healthcheck handler, if that timestamp is less than N seconds old (1 <= N <= 60, I'd suggest 10), then treat that as "compiler pool is reachable". if it's more than N seconds old, then follow the current logic of making an RPC to check.

very-nice-to-have:

backportable to 5.x to allow patching existing Cloud instances without requiring an upgrade to 6.x