Consolidate project metadata

[k-collie]

Description of work

Motivation

• There's both a setup.py and pyproject.toml at the moment, each specifying different dependency constraints
• pip uses the setup.py dependencies when installing from pypi, the unconstrained dependency on Flask results in Flask==3.1.0 being installed, which fails the unit tests. It would be nice to add some constraints when we know some dependency versions are incompatible.
• pip install . uses pyproject.toml dependencies, resulting in different dependency versions when installing from local clone vs from pypi
• Currently server dependencies are always installed, it would be nice if these were optional for client-only users (especially if we also constrain Flask as this could result in conflicts in their environment).

Changes

• Consolidated setup.py and pyproject.toml metadata into pyproject.toml
• Updated requires-python to >=3.9 for new poetry-core version
• Use newer poetry and standard pyproject.toml format
• Made server dependencies into an optional group
  • Will require a major version update since server dependencies are removed by default
• Removed custom VERSION file and used importlib_metadata standard method to expose package version number

Possible further changes

• Can restrict further to Flask<1 if we know Flask>1 doesn't work
  • Tests all passed and coverage was good when ran with Flask==1.1.4, couldn't see any obvious problems in the few uncovered lines
  • Maybe it's job of dependencies to restrict Flask to <1 if they require?
  • Best to keep dependencies loose unless we know it's broken
• Could add up to date maintainers list to pyproject.toml
  • Who should be on this list?
• Removed requirements.txt and requirements_dev.txt
  • poetry.lock should replace requirements_dev.txt sufficiently
  • We could use poetry-export precommit hook to keep a requirements.txt in sync if desired

To test

• Run client and dataclass tests without server optional group installed
• Possibly test with concrete SAL server implementation
• Test client with numpy>=2 installed? No constraint on numpy for pypi installs prior to this MR anyway

Reviewer checklist

• The author suggested tests are successful.
• Relevant documentation (user and developer) has been added, and is clear and complete.
• Suitable unit tests have been added. If unit tests are not included in this MR, please comment as to why and create an issue to address this.