docs: add issue management guideline

[bari12]

Add docs/developer/issue_management.md describing the issue management policy.

[bari12]

Please have a look and send some last comments. I am still drafting a migration plan.

[bari12]

In terms of migration plan (I will bring this up in the rucio meeting as well) I would suggest:

Age-cohort phased migration

• Phase 0 — Preparation (before any closures). Announce the policy and the wishlist concept on Mattermost; create the wishlist and capacity-review-YY/QN labels; enforce the policy on all newly created issues immediately.
• Phase 1 — Issues older than 24 months (~115 issues in main repo, another 30 in the others). These already breach the hard cap, so the default is to close them (to wishlist). Keeping an issue requires an explicit, written justification for that specific issue explaining why it should stay open; any issue without such a justification is closed and labelled wishlist. Decisions are communicated back to the reporter.
• Phase 2 — Issues 12–24 months old. (169 in main repo, 83 in others) Handled at the next regular ticket-review window, where the decision framework is applied as normal (keep / wishlist + close / defer with capacity-review-YY/QN / close). No separate dedicated sweep.
• Phase 3 — Issues under 12 months. A soft approach: these are not reviewed all at once. They are brought into line indirectly over time as they're touched during normal triage and component reviews, and as they naturally age toward the 12-month horizon.

I recognise that this does introduce some temporary unsharpness, by keeping all issues under 12 months in place rather than reviewing them up front, the tracker won't fully reflect the committed plan until that cohort has aged through normal reviews. From what we discussed in the Rucio meeting, I would suggest we accept this as the price of a softer rollout.

[bziemons]

I know these lists have auto-numbering, but if you name "two consequences", I think it makes sense to just number them correctly here so that anyone reading the plaintext version is not confused.