fix(test): seed shard allocations for the storage-attestation e2e

[blacks1ne]

Base: 2b96656e

worker_active_storage_attestation times out after 90s — the shard never produces a frame.

TestProver::to_prover_info builds a ProverInfo with an empty allocations vector. prove_next_state's effective-Active proposing gate asks

info.allocations.iter().any(|a| a.confirmation_filter == self.filter
    && a.effective_status(anchor_gfn) == EffectiveStatus::Active)

which is vacuously false on an empty vector, so every view nullifies with NoVote.

The gate is behind if anchor_gfn > 0. Every other CW test leaves the global anchor at genesis and skips it; this is the only one that seeds a global frame (StorageHarness::seeded(1000)), which is what makes it live. The committee checks upstream of it still pass because TestProverRegistry::get_active_provers ignores the filter and the allocations entirely — so the fixture admits provers to a committee they hold no allocation on, a shape production cannot build.

The test and to_prover_info both date from 961297ec; the gate arrived later in 324ffb3b. It stayed invisible because the E0063 break (#613) stopped the target compiling at all, and CI does not run on this branch (#616).

Fix: seed each member an Active allocation on the shard filter, confirmed for the anchor's epoch.

On epoch: its doc calls 0 a "genesis/grandfather sentinel", but effective_status has no epoch == 0 case. Genesis appears not to need one — genesis provers are seeded with the empty filter, which returns Active before the epoch comparison is reached — so the comment reads as stale rather than as describing missing code. Not touched here.

cargo nextest run -p quil-engine --test e2e_consensus → 14/14 pass; worker_active_storage_attestation 1.229s, previously a 90s timeout.

Out of scope: the gate itself, and the CW/PoRep attestation follow-up noted in the test's own comment.