Enhance delete lifecycle with new hooks (pre/post delete)

[o-orand]

Is your feature request related to a problem? Please describe.
When a bosh delete is executed, it may require additional cleanup (uaa, etc...). Currently, COA require each deployment to develop custom cleanup solution.

Describe the solution you'd like
Enhance delete lifecycle with a pre-delete.sh and a post-delete.sh to ease theses additional cleanup step.
It may impact #67

[gberche-orange]

Limitations of adding shell hooks along with #187 (exposing credhub credentials in shell hooks):

• security risks: limits isolation among deployments/root-deployments
• broadens fragile use of the shell hooks: dependent binaries are not versionned

Alternative is to have deployer-specific life cycle events:

• for bosh: bosh errands invoked as pre-delete steps. Similar to ODB post-deploy and pre-delete life cycle events https://docs.pivotal.io/svc-sdk/odb/0-25/creating.html#service-instance-lifecycle-errands
• for tf: native builtin terraform provisionner which trigger at destroy time.

[gberche-orange]

Impacts if not implementing this enhancement: resources leaks

• backups in S3
• credhub entries
• uaa clients for shield (idempotent creation)

Contexts of these leaks:

• disabled deployment by an operator
• on-demand service instances deployments deleted by end users

[poblin-orange]

also note the new bosh lifecycle https://bosh.io/docs/pre-stop/
bosh deployment could leverage this lifecycle (see eg: orange-cloudfoundry/generic-scripting-release#4)

support added in v3