improve metric `handle_query_duration_seconds` and` tidb_server_query_total`

## Enhancement

we use tidb in a multi-tenant environment, thus it would be helpful to be able to distinguish the caller in some of the monitor metrics, for example, adding a `user` tag in metrics like `handle_query_duration_seconds` and` tidb_server_query_total`, thus we could customize the alerts based on different users.

also, it would be useful that the `result` tag for `tidb_server_query_total` could be more specific. for example, now it has value `OK` and `Error`, it would be useful if the `Error` could be further broken down into server error (like http 5xx) or client error (like http 4xx), so that the alert could distinguish them.