reverse_proxy: add placeholder http.reverse_proxy.retries

[steffenbusch]

The reverse_proxy directive in Caddy already supports a variety of useful placeholders, as documented here

This pull request proposes the addition of a new placeholder: {http.reverse_proxy.retries}. This placeholder tracks the number of retries actually performed to communicate with an upstream. I believe this would be a valuable enhancement, providing more detailed insights into load balancing behavior.

For example, in a Caddyfile, this placeholder might be used with log_append like this:

:8080 {
	log {
		output file /var/log/access.json
	}

	reverse_proxy {
		to localhost:7070
		to localhost:7071
		to localhost:7072

		health_uri /ping
		health_body pong
		health_interval 15s

		lb_policy random
		lb_try_duration 5s
		lb_try_interval 700ms
	}

	log_append * rp_upstream_host "{http.reverse_proxy.upstream.hostport}"
	log_append * rp_upstream_duration_ms "{http.reverse_proxy.upstream.duration_ms}"
	log_append * rp_upstream_latency_ms "{http.reverse_proxy.upstream.latency_ms}"

	log_append * rp_retries "{http.reverse_proxy.retries}"
}

In the access.json, this would then appear like this:

{
  "level": "info",
  "ts": 1725040156.150035,
  "logger": "http.log.access.log1",
  "msg": "handled request",
  "request": {
...
    "tls": {
      "resumed": true,
      "version": "TLSv1.3",
      "cipher_suite": "TLS_AES_128_GCM_SHA256",
      "proto": "h3",
      "server_name": "healthchecks.rocky.stbu.net"
    }
  },
  "bytes_read": 0,
  "user_id": "",
  "duration": 3.890395979,
  "size": 9449,
  "status": 200,
  "resp_headers": {
  ...
  },
  "rp_retries": 3,
  "rp_upstream_latency_ms": 1787.609998,
  "rp_upstream_duration_ms": 1787.975152,
  "rp_upstream_host": "localhost:7072"
}

[mholt]

Thanks; yeah if this works for you I guess we can do this. Just one nit on the naming.