profiles: rename BFTConsensus -> Consensus (BFT is implied, not a flavor)

The function derives the lux/consensus ENGINE sampling params (alpha-of-K snowball:
K/alpha/beta) — Byzantine-safe by construction, so the prefix was noise. Clarified in
the doc that this is the agreement layer, distinct from Quasar (lux/quasar) which seals
each event in a post-quantum QuasarCert on top (the CR's spec.consensus.pqLayers).

profiles/bft_test.go

@@ -5,9 +5,9 @@ package profiles
 
 import "testing"
 
-// TestBFTConsensus_MatchesAvalanche pins the Avalanche-grade derivation: K=min(20,N),
+// TestConsensus_MatchesAvalanche pins the Avalanche-grade derivation: K=min(20,N),
 // alpha=ceil(0.75K) at the 15/20 ratio, BFT-safe at every size.
-func TestBFTConsensus_MatchesAvalanche(t *testing.T) {
+func TestConsensus_MatchesAvalanche(t *testing.T) {
 	cases := []struct {
 		n, wantK, wantAlpha int
 	}{
@@ -20,13 +20,13 @@ func TestBFTConsensus_MatchesAvalanche(t *testing.T) {
 		{50, 20, 15}, // capped at K=20, Avalanche
 	}
 	for _, c := range cases {
-		got := BFTConsensus(c.n)
+		got := Consensus(c.n)
 		if got.SampleSize != c.wantK || got.PreferenceQuorumSize != c.wantAlpha || got.ConfidenceQuorumSize != c.wantAlpha {
-			t.Errorf("BFTConsensus(%d) = K=%d alpha=%d, want K=%d alpha=%d",
+			t.Errorf("Consensus(%d) = K=%d alpha=%d, want K=%d alpha=%d",
 				c.n, got.SampleSize, got.PreferenceQuorumSize, c.wantK, c.wantAlpha)
 		}
 		if !IsByzantineSafe(got.SampleSize, got.PreferenceQuorumSize) {
-			t.Errorf("BFTConsensus(%d) = K=%d alpha=%d is NOT Byzantine-safe", c.n, got.SampleSize, got.PreferenceQuorumSize)
+			t.Errorf("Consensus(%d) = K=%d alpha=%d is NOT Byzantine-safe", c.n, got.SampleSize, got.PreferenceQuorumSize)
 		}
 	}
 }

profiles/production.json

@@ -1,6 +1,6 @@
 {
   "name": "production",
-  "description": "Avalanche-grade BFT consensus for live multi-validator networks. K=5/alpha=4 (80% quorum, >= Avalanche's 15/20 = 75%), Byzantine-safe (2*4-5 = 3 >= 2). For the current 5-validator Lux primary networks; for larger sets derive via profiles.BFTConsensus (K=min(20,N), alpha=ceil(0.75K), e.g. K=20/alpha=15 = exactly Avalanche).",
+  "description": "Avalanche-grade BFT consensus for live multi-validator networks. K=5/alpha=4 (80% quorum, >= Avalanche's 15/20 = 75%), Byzantine-safe (2*4-5 = 3 >= 2). For the current 5-validator Lux primary networks; for larger sets derive via profiles.Consensus (K=min(20,N), alpha=ceil(0.75K), e.g. K=20/alpha=15 = exactly Avalanche).",
   "consensus": {
     "sample-size": 5,
     "preference-quorum-size": 4,

profiles/profiles.go

@@ -64,11 +64,20 @@ type NetworkConfig struct {
 	PingFrequency         string `json:"ping-frequency"`
 }
 
-// BFTConsensus derives Byzantine-fault-tolerant consensus parameters matching
-// Avalanche's safety profile for a validator set of size n. It is the single source
-// of truth for consensus safety: a deploy should derive K/alpha from the LIVE
-// validator count, never hardcode them (the live K=5/alpha=3 drift was sub-BFT —
-// 2*3-5 = 1 < floor((5-1)/3)+1 = 2 — which the consensus engine correctly refuses).
+// Consensus derives the sampling-consensus parameters (the lux/consensus engine's
+// alpha-of-K snowball/snowman knobs: K=sample, alpha=quorum, beta=commit) for a
+// validator set of size n. Consensus is Byzantine-fault-tolerant by construction —
+// there is no non-BFT variant — so the result always satisfies the engine invariant
+// 2*alpha - K >= floor((K-1)/3)+1; "BFT" is implied, not a flavor.
+//
+// This is the ENGINE/agreement layer, NOT post-quantum finality. Quasar (lux/quasar,
+// the per-round QuasarCert + pqLayers in the CR's spec.consensus) seals each agreed
+// event in a PQ weighted certificate ON TOP of this — a separate concern; this struct
+// carries no PQ config.
+//
+// It is the single source of truth for sampling safety: a deploy derives K/alpha from
+// the LIVE validator count, never hardcodes them (the live K=5/alpha=3 drift was
+// sub-BFT — 2*3-5 = 1 < floor((5-1)/3)+1 = 2 — which the engine correctly refuses).
 //
 //   - K (sample size)  = min(n, 20)        — Avalanche caps the poll sample at 20.
 //   - alpha (quorum)   = ceil(0.75*K)      — Avalanche's 15/20 = 75% ratio (K>=4).
@@ -76,10 +85,9 @@ type NetworkConfig struct {
 //                                            the Byzantine floor is used to preserve
 //                                            liveness (one tolerable fault).
 //
-// The result always satisfies the engine invariant 2*alpha - K >= floor((K-1)/3)+1.
 // Examples: n=5 -> K=5,alpha=4 (80%); n=20+ -> K=20,alpha=15 (exactly Avalanche);
 // n=3 -> K=3,alpha=2 (67%, BFT-minimal).
-func BFTConsensus(n int) ConsensusConfig {
+func Consensus(n int) ConsensusConfig {
 	k := n
 	if k < 1 {
 		k = 1