Add Claude memory files for all relevant project files

[Migaroez]

Description

This PR builds on top of #20841 to make using claude code inside the cms code base more efficient.

All relevant projects now have their own memory file with a consise and relevant description of all key aspects of their project.

@iOvergaard Could you have a look at the login project since it's outside my area of expertise
@Zeegaan Could you have a look at the hybrid cache one, I couldn't see anything wrong at first glance but I trust your expertise more than mine.

[iOvergaard]

Very impressive summary of the Login project. I cannot find any faults in its description. LGTM.

[AndyButland]

Just a question - there's a tonne of detail in here around file counts, file names, file locations etc., that will go out of date.

What's the process for keeping these up to date? I assume not manually! Do we need a to agree a process and have a scheduled period when we regenerate them? Or is this something to do on each PR?

[Zeegaan]

@AndyButland I was just thinking the same thing when reading this through! 🤣
The maintenance of these will be a big headache unless we can do something.
In my opinion they have to be updated on each PR, as we should minimize working with out of date files. We might run into something like Human error from a merge conflict aswell 🙈
Maybe we could do something like you suggested, eg. regenerating them / getting an AI assistant to update them on each PR.

The hybrid cache file looks good though ✅

[iOvergaard]

I think you should see the list of files more as a pattern than an actual list of files. The AI will use that to figure out where similar files should go. But we can optimize along the way and make it read more as an actual pattern than the current static structure.

[Migaroez]

The current conclusion (after talking with Phil and him talking with Matt B) regarding the maintance of references in the memory files is (TLDR: we will handle this as Agile as possible):

• We don't know how severe the impact is when they change, claude might just update its context by paging up or down in the file a bit or it might get a bit confused and quote the wrong line. We can and will test this at some point (If Phil doesn't get to it ill probably do it in the next cycle hack). Please notify us if you see this happening naturally.
• Whether there is an impact or not, we should probably keep them up to date for human readability anyway, so at a minimum we should verify/update them every major
• Depending on the severity we can update them periodically dependabot style with a human verification of the generated PR
• The current commands of generating the files already take regenerating into account and can be used to update the reference, but it can also add things that were removed during the initial creating review process.
• We can fine tune a prompt (turned into a command) to create an as minimal as possible PR to speedup this process.