Cloudwanderer

- Collapsed all `write_secondary_attributes` methods into `write_resources` so secondary attributes are written automatically.
- Moved AWS specific methods to `CloudWandererBoto3Interface`
- Fixed bug that would have prevented global services with regional resources being cleaned properly.
This was due to cleanup only happening in the global service region, and being limited to
cleaning up resources in that region. E.g. it would write buckets from all regions from `us-east-1`
and then _only_ cleanup `us-east-1` s3 buckets.
- Removed `client_args` as an explicit argument on cloudwanderer resources, any keywords args supplied to `write_` methods are now passed into the `get_` methods of the `cloud_interface`
- Subresources now build their compound id using a `/` separator rather than a `:` separator. This ensures that
`:` remains the primary separator for URN parts.

- Reuse service definition objects via `_get_resource_definitions` to save on time spent reinstantiating identical objects
- Only instantiate the default `ServiceMapping` from `get_service_mapping` if it's required.
- Reduced number of regions tested from _all_ to 3.

- S3 buckets and other regional resources of global services will now only be written in their service's globalRegion
- Fixed bug where AWSUrn did not parse subresources URNs correctly
- Improved cloudwanderer tests by leveraging `MemoryStorageConnector`

- Added `role_managed_policy_attachments` secondary attribute
- Added `role_inline_policy_attachments` secondary attribute
- Collapsed secondary attribute definitions into custom resource definitions
- Collapsed boto3 resources into custom resource definitions
- Abstracted GlobalServiceMaps to ServiceMaps as supporting metadata CloudWanderer needs to understand resources
- Resources with multple identifiers now include all identifiers as part of the AWSUrn
- Split out experimental concurrency into separate `write_resources_concurrently` method.

- Added `get_secondary_attribute` to `CloudWandererResource`
- Added `is_inflated` to `CloudWandererResource`
- Storage Standardisation improvements
- Standardised storage connector write tests
- Deleted duplicate read tests
- Added `write_resource_attribute` to `BaseStorageConnector`
- Added `write_resource_attribute` to `MemoryStorageConnector`
- Renamed `resource_attributes` to `secondary_attributes`
- Fixed bug where `write_secondary_attributes` would enumerate services which did not have secondary attributes.

- Added support for multiple storage connectors
- Made Storage Connectors the primary interface for reading from storage
- `read_resource` on DynamoDbStorageConnector returns a `CloudWandererResource` instead of an iterator.
- `read_resource` on MemoryStorageConnector returns a `CloudWandererResource` instead of an iterator.
- Added `read_resources` to DynamoDbStorageConnector
- Added `read_resources` to MemoryStorageConnector
- Removed `read_resource_of_type` from DynamoDBStorageConnector
- Removed `read_resource_of_type_in_account` from DynamoDBStorageConnector
- Removed `read_all_resources_in_account` from DynamoDBStorageConnector
- Removed `read_resource_of_type` from BaseStorageConnector
- Removed `read_resource_of_type_in_account` from BaseStorageConnector
- Removed `read_all_resources_in_account` from BaseStorageConnector
- Removed `read_resource_of_type` from MemoryStorageConnector
- Removed `read_resource_of_type_in_account` from MemoryStorageConnector
- Removed `read_all_resources_in_account` from MemoryStorageConnector
- Removed `read_resource_of_type` from CloudWanderer
- Removed `read_resource` from CloudWanderer
- Removed `read_all_resources_in_account` from CloudWanderer
- Removed `read_resource_of_type_in_account` from CloudWanderer
- Added dynamodb filter expressions to increase flexibility of `get_resources`
- Added attribute projections for urn parts to DynamoDB Global Secondary Indexes