✨ Feature request: Multi-target scraping via /probe endpoint

Background

Today each elasticsearch_exporter instance can scrape one cluster – the one given in --es.uri (or default http://localhost:9200).
In environments with dozens of SaaS-managed or shared clusters this forces operators to run one exporter per cluster, which is operationally heavy.

Proposal

1. New HTTP endpoint /probe

   • Called with ?target=<http[s]://host:port> (and optional auth_module=).
   • On each request the exporter spins up a fresh Prometheus Registry, registers the usual collectors with the supplied target URL and streams the metrics.

2. Auth modules (config file)

   auth_modules:
     prod_basic:
       type: userpass
       userpass:
         username: metrics
         password: s3cr3t
       options:
         sslmode: disable
   
     staging_key:
       type: apikey
       apikey: BASE64IDKEY==

   – File is loaded once at startup with --config.file=.
   – /probe?target=...&auth_module=staging_key injects the creds/header.

3. Alias on /metrics
   If a request to /metrics already contains target= it is internally routed to the probe handler, preserving full backward compatibility while letting users keep path: /metrics in ServiceMonitors.

4. No breaking changes
   • Classic single-target /metrics behaviour is untouched.
   • All existing CLI flags (--es.indices, TLS flags, AWS signing, etc.) still apply.

Example Prometheus config

- job_name: es-multi
  metrics_path: /probe          # or /metrics
  params:
    auth_module: [prod_basic]   # default auth module
  static_configs:
    - targets:
        - https://es-prod:9200
    - targets:
        - https://es-staging:9200
      labels:
        __param_auth_module: staging_key
  relabel_configs:
    - source_labels: [__address__]
      target_label: __param_target
    - source_labels: [__param_target]
      target_label: instance
    - target_label: __address__
      replacement: exporter:9114   # exporter service

Acceptance criteria

• /metrics continues to expose the startup cluster.
• /probe?target=… returns metrics for any reachable cluster.
• Auth-module userpass and apikey work as expected.
• Alias /metrics?target=… behaves identical to /probe.
• Documentation & example YAML updated.

Feedback and additional requirements welcome!