low interactivity of `create table like hive` and yarn mode

Test on 0.7.2. Test again on main 1317ce70, if need fix, create a task

1.

- [ ] Get `reached timeout` error, cuz taskmanager don't response for `create hive`. `create hive` job timeout is from variable `job_timeout`, default 20s, enough in yarn mode? (not 60s in sql_cluster_router, global var in nameserver) will be fixed with i3342

2. We can't get the SPARK_HOME in effect directly, need to ask user more about it. It's better to get runtime config from all severs.
3. If Spark can't work, e.g. spark_home is wrong, or spark-submit is bad, we only get `reached timeout` error. We can test it by diag tool. Or we should give user a simple sql script to test as an alternative(in must read doc and install_deploy doc).
4. And `show joblog` can't work, `reached timeout` too. No useful error msg. Why show joblog failed(just read a file in local)?
5. yarn mode document about spark-submit version, and spark version.
6. `create table like hive` will run in no used db, track it on #3149

The issue mentions problems with 'create table like hive' in yarn mode, including timeouts, SPARK_HOME configuration, and joblog failures. Look at the code for job timeout settings (job_timeout, sql_cluster_router) and the spark-submit integration. Check the diagnostic tool and documentation for yarn mode. The issue is tracked with other related tasks (#3149, i3342).

Rédigé par le modèle d'indexation à partir du texte de l'issue.