node: Expect only Postgres host names for GRAPH_POSTGRES_SECONDARY_HOSTS

server: Rename the metrics for query execution for consistency

We rename the two metrics from subgraph_query_execution_time and
subgraph_failed_query_execution_time to query_execution_time and
query_failed_execution_time so that their Prometheus namespace is 'query'
which sets it apart from the 'deployment' metrics which are all about
indexing.

We also rename the label that connects measurements to a deployment from
'subgraph_deployment' to 'deployment' since that's what the other metrics
use

node, runtime/wasm: Fix tests

all: Address review comments for load management

Comments are in Zac's review of
graphprotocol#1762