refactor(Photoshop): Restructure logic that establishes a Dispatch object

[Investigamer]

Motivation for this Commit

After the popularity of Photoshop Beta due to the Generative Fill feature, as well as the Photoshop version rollover from v24 to v25, there has been increased occurrence of strange issues such as:

• Failing to establish a valid Dispatch object, despite Photoshop being installed
• Resolving Dispatch objects that conflict with one another, likely due to a version conflict i.e. having two versions of Photoshop installed, or having lingering registry data from an uninstalled version of Photoshop.
• Other issues I've been unable to diagnose but we've seen on multiple users' systems. Some of these probably stem from "cracked" Photoshop installations that improperly install registry data, however I've personally seen cases where the registry entries appear to be fine and issues persist.
• Some of these issues remain a mystery and require further investigation, some may be unfixable (for example, entirely missing registry data due to a portable or poorly designed "cracked" Photoshop installation, these are not our responsibility).

Goals of this Commit

• Restructure the sequence of events that attempt to resolve a valid Dispatch object in Photoshop.__init__(), ensuring the sequence is optimized. My current solution:
  • Check if a Photoshop version was provided manually. If valid, attempt to resolve a Dispatch object using this version.
  • Pull a list of version ID strings found in registry data (e.g. 170, 180, etc), add a final fallback case of empty string.
  • For each version ID (in descending order) attempt to resolve a Dispatch object, with the last attempt being an empty version (e.g. Photoshop.Application).
  • Finally, if all attempts to resolve a Dispatch object fail, raise a PhotoshopPythonAPIError.
• Implement an internal logger for providing optional debug warnings, which can be toggled with environment variable 'PS_DEBUG' ahead of time.
• Refactor and simplify code for readability, make small adjustments to improve efficiency.
• Adjust the nomenclature of a couple internal attributes and methods.
• Reinforce docstrings and type declarations.

Note on Limitations

These changes surely don't address every Dispatch related issue we've seen, and unfortunately I cannot replicate many of these reported issues on a local system, so its impossible for me to test each scenario. Nevertheless, I put in a few days of deep dive research into the topic, and I'm confident my proposed methodology is a step in the right direction.

[Investigamer]

Additional Commit: application.py

• Wrap failed app.colorSettings setter call with human readable exception
• Allow app.load() to receive str or Path
• Fix app.purge() docstring to match definitions in PurgeTarget
• Reinforce type declarations and improve docstring grammar

[Investigamer]

Additional Commit: Deprecate Python 3.7

• I recommend we move to deprecate support of Python 3.7 since it reached End of Life earlier this year. Currently my commit is not compatible with Python 3.7 since @cached_property wasn't implemented until 3.8. I went ahead and removed Python 3.7 from our workflows and project files, let me know if you prefer to try and maintain support for 3.7 and I will reverse this change and find a replacement for my @cached_property solution. 👍
• Fixed an error when trying to publish docs with mkdocs, appears to be caused by an outdated version of importlib_metadata.
• Added support for Python 3.12 to workflows and project files. Python 3.12 appears to work just fine with the API on my system and has passed the import test workflow!
• Fixed circular import error