Add external schema support

[sworisbreathing]

This PR extends the redshift_schema resource and data source to support all of the available flavours of external schema documented at https://docs.aws.amazon.com/redshift/latest/dg/r_CREATE_EXTERNAL_SCHEMA.html. There is a new nested configuration block in the redshift_schema resource, named external_schema.

Supported external schema sources are:

• data_catalog_source for AWS Glue Data Catalog
• hive_metastore_source for a Hive metastore (I think really it only supports EMR but I haven't tested it as I don't have a Hadoop environment to test against, EMR or otherwise)
• rds_postgres_source for federated query against RDS/Aurora PostgreSQL
• rds_mysql_source for federated query against RDS/Aurora MySQL (note that this is currently a Preview feature and thus subject to change. Your Redshift cluster must be on the Preview maintenance track to use this)
• redshift_source for querying a datashare from another redshift cluster (both clusters must use an instance family which supports data sharing... currently only the new RA3 instance types support data sharing)

Destroy/recreate is forced when updating a terraform-managed schema, if:

• you add external_schema to an existing redshift schema resource (switch the schema from internal to external), or
• you remove external_schema from an existing redshift schema resource (switch the schema from external to internal), or
• you change any of the settings in the external_schema block

This is required because ALTER SCHEMA does not support external schemas (except for changing the owner)

I have included acceptance tests for all data sources, however they are conditionally enabled on specific environment variables being set. I have successfully run the ones for data_catalog_source, rds_postgres_source, and redshift_source.

[sworisbreathing]

Just a quick note on this one for the sake of having a written record somewhere... AWS support have confirmed that svv_external_schemas.esoptions will be truncated to 256 characters whenever you join on svv_schema_quota_state. Joining on svv_schema_quota_state turns the query into a "padb query", and padb queries treat text columns as varchar(256). They haven't really made it clear to me why it's a padb query but they said that's the reason.

So, if anyone is looking to do make changes around the schema resource/data source, please be sure that the query on esoptions doesn't have any joins in it, otherwise you'll risk introducing a regression which prevents the provider from reading back the external schema configuration.