Fix: Preserve Elasticsearch URL path in stats requests

[Robert-Steiner]

Description

When the Elasticsearch URL contains a path (e.g., http://example.com/elasticsearch/), the path is incorrectly stripped when making _all/_stats requests, resulting in requests to http://example.com/_all/_stats instead of http://example.com/elasticsearch/_all/_stats.

The regression was introduced in v1.10.0 when the URL resolution logic was changed from using path.Join to url.ResolveReference.

package main

import (
    "fmt"
    "net/url"
)

func main() {
    base, _ := url.Parse("http://example.com/elasticsearch/")
    fmt.Println(base.ResolveReference(&url.URL{Path: "/_all/_stats"}))
    fmt.Println(base.ResolveReference(&url.URL{Path: "_all/_stats"}))
}

Output

http://example.com/_all/_stats
http://example.com/elasticsearch/_all/_stats

This PR fixes the URL path handling to ensure custom paths in the Elasticsearch URL are properly preserved when constructing API endpoint requests.

[ArthurSens]

Hi @Robert-Steiner, sorry for the delay here. Looks like the PR needs a rebase now.

Also, could you add a test to prevent us from re-introducing this bug again in the future?

[ArthurSens]

Maybe we should be using i.url.JoinPath here?

[sysadmind]

Based on my research, I think JoinPath should become our standard. It covers the prefix needs we have.