Add env var to skip initdb

[JacobHayes]

I'm trying to build a data container to ship developer databases (up to date migrations, fixture data, etc which are managed elsewhere). We could use a docker-entrypoint-initdb.d script to move a bundled data folder to PGDATA (similar to what's suggested in #105), but then the entire initdb process was a (one time) waste.

I'm proposing a new USE_SEED_DB env var that would skip initdb and instead move everything in a "seed" directory to PGDATA. This would set the expectation that scripts in docker-entrypoint-initdb.d/ would act on the seed db and not a fresh one.

I'm willing to submit a PR for this.

[JacobHayes]

We could use /docker-entrypoint-seed-db.d for example.

[JacobHayes]

I've opted to use a data container based on scratch (or rather tianon/true since docker-compose requires a command).

Pro: image can be used without running (used to need to run docker-entrypoint.sh to set things up)

Con: mounted volume directory may not have 700 permissions (must being careful when creating data image).

I've submitted #110 to set 700 explicitly.