Actually record last retry time after reconnecting

The code was incorrectly setting it to "never", meaning the
bodyReaderMSSinceLastRetry heuristic was never triggered,
and, in effect, we would _always_ reconnect on io.ErrUnexpectedEOF
or syscall.ECONNRESET .

This corrects the code - but it might also make more network failures
user-visible instead of silently retried.

Signed-off-by: Miloslav Trmač <[email protected]>

docker/body_reader.go

@@ -197,7 +197,7 @@ func (br *bodyReader) Read(p []byte) (int, error) {
 		consumedBody = true
 		br.body = res.Body
 		br.lastRetryOffset = br.offset
-		br.lastRetryTime = time.Time{}
+		br.lastRetryTime = time.Now()
 		return n, nil
 
 	default: