S3Metadata should lazily load schemas

[adrdc]

in ctor for S3Metadata object (ConnectorMetadata), it gets table config. that process involves looking for metadata search keys from OBS/ECS, requesting data from schema registry, loading schemas from disk, etc.

however this is only needed on the coordinator. this info is not accessed on workers. so N workers will be pulling this same info down all to just be not used.

@Inject
public S3Metadata(S3ConnectorId connectorId,
                  Supplier<Map<SchemaTableName,S3Table>> s3TableDescriptionSupplier,
                  S3ConnectorConfig s3ConnectorConfig) {
	...
        this.tableDescriptions = s3TableDescriptionSupplier.get();
        ...