docs(phase-0): propose pre-cell performance + security patches

[aleid1]

Summary

Proposes five Django write-path patches + one AWS credential rotation that should ship before any cell-based sharding work. All independent, all feature-flagged, all reversible.

What this PR is

• docs/phase-0-patches-proposal.md — one proposal doc. Nothing else.
• No code change. No schema change. Safe to merge at any time (or to close).

What it is not

• This is not the code PR. Once we agree on a spec here, a separate PR introduces the Python changes + tests per patch.

Why

Study in nelc/openedx-2.0 identified five Django-level write patterns as the root cause of MySQL contention we see on moe.futurex.sa and a plaintext AWS key in edx-platform-strains/teak/nelp/config.yml. The proposal doc names exact files, exact flags, exact rollback for each.

Patches proposed

1. Defer auth_user.last_login to Redis
2. Async + debounced BlockCompletion.submit_completion
3. eox_nelp likedislikeunit 404-storm fix
4. CONN_MAX_AGE=60 + CONN_HEALTH_CHECKS=True
5. Route eox-nelp signal tasks to event_routing queue
6. Rotate the plaintext AWS keys in edx-platform-strains

Ordering and verification commands are in the doc.

Test plan

• Team review of each patch spec in the doc (inline comments preferred).
• For each approved patch, open a separate PR with code + unit tests + a staging soak plan.
• Do not merge this PR until at least one patch below has a follow-up code PR open, so the reference stays live.

Related

• PRD + evidence: https://github.com/nelc/openedx-2.0
• Full spec source: https://github.com/nelc/openedx-2.0/tree/main/django-patches