docs: Alignment analysis of .NET and Python context compaction vs. ADR-0019

[Copilot]

Motivation and Context

Adds a cross-implementation analysis of the in-progress Python (#4469) and .NET (#4496) compaction PRs against ADR-0019 to surface alignment gaps before either PR merges.

Description

Creates docs/design/alignment.md — no code changes.

Spec recap (ADR-0019 decision)

• Option 1 (standalone CompactionStrategy) + Variant F2 (_-annotated messages, no sidecar)
• Three required compaction points: in-run, pre-write, existing-storage
• TokenBudgetComposedStrategy as the canonical composition pattern
• Strategy-internal trigger guard (no separate trigger abstraction)

Python PR #4469

• ✅ Closely follows spec: F2 model, Protocol interface, TokenBudgetComposedStrategy, strategy-internal triggers, all four built-in strategies
• ⏳ Pre-write and existing-storage compaction deferred to Phase 2
• ⚠️ Reasoning-message atomicity not yet handled in annotate_message_groups()

.NET PR #4496

• ✅ Sound in-run compaction: CompactingChatClient (innermost in pipeline), MessageIndex with incremental Update(), state serialized into AgentSession.StateBag, streaming support
• ⚠️ Uses F1 (sidecar MessageIndex) rather than spec-preferred F2 — intentional and acceptable per spec's own acknowledgement
• ❌ TokenBudgetComposedStrategy absent; PipelineCompactionStrategy runs all strategies unconditionally with no budget-stop
• ⚠️ Pre-write compaction via IChatReducer on InMemoryChatHistoryProvider — separate abstraction from CompactionStrategy, not unified as the spec intends
• ❌ Reasoning-message atomicity not handled in MessageIndex.Create

Key gaps identified

Priority	Gap	Affects
High	TokenBudgetCompactionStrategy missing	.NET
High	Pre-write uses IChatReducer, not CompactionStrategy	.NET
High	Pre-write/storage compaction not yet implemented	Python (Phase 2)
Medium	Reasoning-message (ReasoningContent) atomicity	Both

Contribution Checklist

• The code builds clean without any errors or warnings
• The PR follows the Contribution Guidelines
• All unit tests pass, and I have added new tests where possible
• Is this a breaking change? If yes, add "[BREAKING]" prefix to the title of the PR.

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.