When recovering from an URL, only continue without failing if
checkpoint data request fails with 404 code. If the request fails
with other 4xx HTTP error code, go-spacemesh startup fails. Upon
other errors, including 5xx HTTP errors and network errors, retry
checkpoint data request several times.
`recovery.retry-count` and `recovery.retry-interval` specify retry
parameters, defaulting to 5 and 3s, respectively.

When this flag is specified, go-spacemesh doesn't exit when recovery
URI request fails. Retries are still done, though.
This is needed b/c in the context of TestRecovery systest, all the
nodes are created with --recovery-uri flag passed, including the
bootnodes. This way, they can't successfully start unless the
bootstrapper is already running. But the bootstrapper requires an
working go-spacemesh node to start up, thus we get a circular
dependency.