Airflow scheduler with statsd enabled crashes when dag_id contains unexpected characters

### Apache Airflow version 2.1.1 ### Operating System Oracle Linux 7.9 ### Versions of Apache Airflow Providers _No response_ ### Deployment Virtualenv installation ### Deployment details The problem initially happened in a virtualenv installation of airflow 2.1.1 running as a systemd service. I tried to reproduce it using docker-compose and airflow version 2.1.0; the problem occurs there as well. ### What happened A new DAG was added with the symbol "ö" in its dag_id. As soon as the DAG was triggered, the airflow scheduler died unexpectedly and could not be started again until the DAG was deleted from the UI and its dag_id changed in the DAG file. ### What you expected to happen The scheduler should continue running even if an exception occured when emitting metrics for a DAG run. An exception related to a single DAG run should not be able to kill the entire scheduler process. Ideally, if a dag_id contains characters which are not allowed, the error should appear as soon as Airflow tries to parse it, and it should not be possible to schedule it at all. It would also be nice if the error message showing all allowed characters displayed them in alphabetic order. ### How to reproduce 1. Start an instance of airflow with statsd metrics enabled 2. Create a DAG with a dag_id containing symbols which are not allowed 3. Trigger the DAG. The scheduler will shut down as soon as the DAG is triggered. ### Anything else The problem occurs only when AIRFLOW__METRICS__STATSD_ON=true . When statsd metrics are disabled, the DAG runs without problems. Scheduler output with some values replaced with equivalent placeholders: ``` Sep 03 14:25:37 my.airflow.domain python[110341]: [2021-09-03 14:25:37,134] {{dagrun.py:444}} INFO - Marking run <DagRun other_dag @ 2021-09-03 11:24:34.7 60370+00:00: scheduled__2021-09-03T11:24:34.760370+00:00, externally triggered: False> successful Sep 03 14:25:37 my.airflow.domain python[110341]: [2021-09-03 14:25:37,147] {{scheduler_job.py:1229}} INFO - …