feat(prometheus_scrape source): add scrape_delay to stagger scrapes

changelog.d/prometheus_scrape_scrape_delay.feature.md

@@ -0,0 +1,5 @@
+The `prometheus_scrape` source has a new `scrape_delay` option that controls when scrapes happen relative to the configured `scrape_interval_secs`. Sources sharing an interval otherwise all scrape at the same instant, so an instance running many of them raises load in a single spike each interval.
+
+`none`, the default, keeps the existing behavior unchanged. A delay in seconds, such as `30s`, holds the first scrape back by that much and then keeps the same fixed cadence, so sources can be staggered by hand. `auto` instead picks a position inside each interval, derived from a hash of the host name, the component ID and the scrape number, which reduces persistent alignment with other periodic work. See the option docs for the trade-offs `auto` makes.
+
+authors: taloric

src/sources/http_client/client.rs

@@ -374,6 +374,8 @@ impl SourceConfig for HttpClientConfig {
         let inputs = GenericHttpClientInputs {
             urls,
             interval: self.interval,
+            initial_delay: Duration::ZERO,
+            jitter_seed: None,
             timeout: self.timeout,
             headers: self.headers.clone(),
             content_type,