feat(document): improve proxy status polling resilience

[joshtrichards]

• Resolves: # (various, possibly)
• Target version: main

Summary

• Replace the unbounded 100ms polling loop for proxy.php status checks with limited exponential retries.
• Split server-starting retries from transport failures so we can show a more specific fallback error when the built-in CODE server cannot be reached at all.

This avoids indefinite waiting and improves startup robustness / user feedback for built-in server deployments.

TODO

• Adjust tests

Notes

Possible future follow-up:

• unify/extract out shared proxy-status logic
  • we have two proxy status implementations: one in src/document.js and one in src/services/collabora.js (and Office.vue indirectly which uses the latter basically).

Checklist

• Code is properly formatted
• Sign-off message is added to all commits
• Documentation (manuals or wiki) has been updated or is not required

[elzody]

@joshtrichards Is there anything else left to finish up here?

[joshtrichards]

Tests at the very least.

When I was working on this last, I'd simultaneously started trying to get a CODE integration test added in the CI chain (in a separate draft PR).

I think I got distracted on that. ;-)

There may be some other things in this PR itself still... I'll have to look over my notes. I have a window that's been sitting open on my desktop specifically to revisit/wrap this... So there was probably something else.

I'll try to check this coming week!