fix: close retried responses in retry middleware

[umputun]

Retry kept a response it decided to retry in lastResponse and looped, and nothing closed its body:

if !r.shouldRetry(resp) {
	return resp, nil
}

lastResponse = resp // next attempt overwrites it, body still open

net/http releases a connection only after the body is read to EOF or closed, so each retried response held one TCP connection and its transport goroutines for the lifetime of the client. The reproducer from #15 (a server answering 502, 502, 200, five requests) ends with 11 connections still open and 35 goroutines.

RoundTrip now closes the body of a response it drops in favor of another attempt, both for a retryable status and for a response returned together with an error. The response of the last attempt is returned to the caller open, as before.

The body is closed without draining, which differs from the fix suggested in the issue. A read in the middleware has no time bound, so a server that sends headers and then stalls the body would block the retry forever on a client without a timeout, where today Retry moves on to the next attempt. Closing alone is enough to release the connection. Same reproducer after the change:

	connections opened	still open	goroutines
before, go1.25.0 and go1.27.1	11	11	35
after, go1.27.1	1	1	5
after, go1.25.0	11	1	5

The go1.27.1 transport drains on close within its own limits and reuses the connection. The go1.25.0 transport does not, so there a retried response costs a new connection, which is closed and no longer leaked.

Related to #15

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.