feat(cost): compare streaming summary lifecycles

[zzylol]

Why

A streaming resource estimate is not enough to choose whether one summary state should be ephemeral, prepared, shared, or continuously maintained. Those alternatives have different legality constraints and different build, update, retention, read, and retirement schedules.

What

Connect analytical streaming resource estimates to the existing summary-maintenance lifecycle alternatives without defining a second lifecycle domain model.

How

• Convert build, update, read, retention, and retirement resources into lifecycle cost inputs for one summary state.
• Apply the existing lifecycle legality and runtime-capability checks before ranking.
• Evaluate alternatives over the same workload horizon and recurrence evidence.
• Leave missing resource evidence unavailable rather than substituting zero.

Before this PR

Streaming resource dimensions could be estimated, but lifecycle alternatives were not compared with those estimates.

After this PR

Each legal lifecycle alternative for an individual summary state can be compared over one horizon using all evidence required by that lifecycle estimate. This PR does not yet claim that a nested SummaryExpr DAG is complete; #335 adds recursive whole-DAG accounting and shared-node deduplication.

Verification

• cargo test --workspace
• cargo clippy --workspace --all-targets --all-features -- -D warnings
• cargo fmt --all -- --check

Stack

Depends on #331. #335 adds complete SummaryExpr DAG estimation.