Many-to-Many relationships cannot be cleaned up inside beforeDelete()

### Task List

- [x] Steps to reproduce provided
- [x] Stacktrace (if present) provided
- [x] Example that reproduces the problem uploaded to Github
- [x] Full description of the issue provided (see below)

### Steps to Reproduce

1. Create domain classes with a many-to-many relationship
2. Attempt to clean up the generated join table inside of the `beforeDelete()` method (utilizing `withNewSession` as outlined here: http://gorm.grails.org/6.0.x/hibernate/manual/#_the_beforedelete_event)
3. Attempt to delete one side of the relationship

### Expected Behaviour

The `beforeDelete()` method is able to remove the object being deleted from other associations in the database.

### Actual Behaviour

An exception is thrown when the session is flushed:

```
Hibernate operation: could not execute statement; SQL [n/a]; Referential integrity constraint violation: "FKGHKKY8WMH379RPMFH92T807RY: PUBLIC.TEAM_MEMBERS FOREIGN KEY(PERSON_ID) REFERENCES PUBLIC.PERSON(ID) (1)"; SQL statement:
delete from person where id=? and version=? [23503-194]; nested exception is org.h2.jdbc.JdbcSQLException: Referential integrity constraint violation: "FKGHKKY8WMH379RPMFH92T807RY: PUBLIC.TEAM_MEMBERS FOREIGN KEY(PERSON_ID) REFERENCES PUBLIC.PERSON(ID) (1)"; SQL statement:
delete from person where id=? and version=? [23503-194]
```

### Notes

Following the advice from [GORM Gotchas (Part 2)](https://spring.io/blog/2010/07/02/gorm-gotchas-part-2/) I understand that many-to-many relationships do not cascade deletions, so I need to manage them on my own. The domain's `beforeDelete()` method seemed like the perfect place to document this and other side-effects of deleting an instance.

If this is not possible I think there should be a section in the documentation that explains a recommended way of cleaning up associated join tables for many-to-many relationships. I think this situation is common enough to warrant it. I don't mind writing that section if it's confirmed that the sample project exhibits correct behavior.

### Environment Information

- **Operating System**: Ubuntu MATE 17.10
- **GORM Version:** 6.1.8.RELEASE, 6.0.13.RELEASE
- **Grails Version (if using Grails):** 3.2.12
- **JDK Version:** `openjdk version "1.8.0_162"`

### Example Application

- https://github.com/joemccall86/cascade-delete-test/tree/automatic-collection-purge
- Specifically the `automatic-collection-purge` branch

Start with the linked example application's automatic-collection-purge branch and reproduce the failure through the domain beforeDelete() method. Read the referenced GORM beforeDelete documentation and inspect the flush-time referential-integrity exception. Done means the many-to-many join-table cleanup behavior is corrected or a documented recommended cleanup approach is established.

Written by the indexing model from the issue text.