monitor: Fix monitor race condition doing hypervisor.check()

The thread monitor will check if the agent and the VMM are alive every
second in a blocking thread. The Cloud hypervisor API server is
single-threaded, if the monitor does a `check()`, while a slow request
is still in progress, the monitor check() method will timeout. The
monitor thread will stop all the shim-v2 execution.

This commit modifies the monitor thread to make it check the status of
the hypervisor after 5 seconds. Additionally, the `check()` method from
cloud-hypervisor will use the method `clh.isClhRunning(timeout)` with a
10 seconds timeout. The monitor function does no timeout, so even if
`hypervisor.check()` takes more 10 seconds, the isClhRunning method
handles errors doing a VmmPing and retry in case of errors until the
timeout is reached.

Reduce the time to the next check to 5 should not affect any functionality,
but it will reduce the overhead polling the hypervisor.

Fixes: #2777

Signed-off-by: Carlos Venegas <jose.carlos.venegas.munoz@intel.com>

src/runtime/virtcontainers/monitor.go

@@ -14,7 +14,7 @@ import (
 )
 
 const (
-	defaultCheckInterval = 1 * time.Second
+	defaultCheckInterval = 5 * time.Second
 	watcherChannelSize   = 128
 )