Octue

Summary
Fix the installation of `octue` when including the `dataflow` extra by pre-determining the packages needed for `apache-beam` to work with GCP. This also significantly slims down the dependencies of `octue` by only including those explicitly needed.

<!--- SKIP AUTOGENERATED NOTES --->
Contents ([495](https://github.com/octue/octue-sdk-python/pull/495))

Dependencies
- Widen range of acceptable `packaging` versions to avoid conflict with `poetry`
- Remove `gcp` option from `apache-beam` package
- Explicitly add necessary GCP package ranges for Dataflow

Fixes
- Ensure `octue.services` namespace is used for Dataflow service topics

<!--- END AUTOGENERATED NOTES --->

<!--- SKIP AUTOGENERATED NOTES --->
Contents ([499](https://github.com/octue/octue-sdk-python/pull/499))

Operations
- Use `python3.9` for testing
- Use and run latest pre-commit hooks

Dependencies
- Align `pyproject.toml` code style package versions with those used by `pre-commit`
- Replace `flake8-isort` with `isort`

<!--- END AUTOGENERATED NOTES --->

<!--- SKIP AUTOGENERATED NOTES --->
Contents ([468](https://github.com/octue/octue-sdk-python/pull/468))

Enhancements
- Call `Analysis.finalise` if it isn't called in an app

Fixes
- Ensure output values sent to `stdout` by CLI are JSON formatted

<!--- END AUTOGENERATED NOTES --->

Summary
Avoid potential race conditions between file and metadata upload in, for example, Cloud Functions that are triggered by file upload and rely on certain metadata being immediately present on those files.

This potential problem was brought to light in this issue: https://github.com/aerosense-ai/data-gateway/issues/87.

<!--- SKIP AUTOGENERATED NOTES --->
Contents ([494](https://github.com/octue/octue-sdk-python/pull/494))

Fixes
- Add metadata to blobs before uploading to cloud storage

<!--- END AUTOGENERATED NOTES --->

Summary
Use signed URLs to point to cloud datasets in input manifests when sending them to children for processing. This means the children will always be able to access the input datasets - not just in the special case of the parent and child having access to the same buckets.

<!--- SKIP AUTOGENERATED NOTES --->
Contents ([491](https://github.com/octue/octue-sdk-python/pull/491))

Fixes
- Use signed URLs when sending input manifests to services

Refactoring
- Factor out dataset URL signing to new `Manifest` method

Testing
- Simplify service tests' arguments
- Update test method docstring
- Ignore irrelevant files in test
<!--- END AUTOGENERATED NOTES --->

Summary
Allow metadata from datafiles instantiated from signed URLs to be retrieved. This has been tested with the real Google Cloud Storage endpoint and works in production, but doesn't currently with the storage emulator used in our tests. See 489 for more information.

<!--- SKIP AUTOGENERATED NOTES --->
Contents ([490](https://github.com/octue/octue-sdk-python/pull/490))

Fixes
- Get datafile metadata when instantiating from a signed URL

Dependencies
- Use latest `gcp-storage-emulator` for tests
<!--- END AUTOGENERATED NOTES --->